Université de Montréal
Long-Term Economic Decline and Non-Mainstream Voting
par
Andrea M.L. Perrella
Département de science politique
Faculté des arts et des sciences
Thèse présentée à la Faculté des études supérieures en vue de l’obtention du grade de Philosophiae Doctor (Ph.D.)
en science politique
Novembre 2005
© Andrea ML. Perrella, 2005 jradeccnfé’\ àcompterdu s ç C ) D > Université de Montréal Direction des bibliothèques
AVIS
L’auteur a autorisé l’Université de Montréal à reproduite et diffuser, en totalité ou en partie, pat quelque moyen que ce soit et sur quelque support que ce soit, et exclusivement à des fins non lucratives d’enseignement et de recherche, des copies de ce mémoire ou de cette thèse.
L’auteur et les coauteurs le cas échéant conservent la propriété du droit d’auteur et des droits moraux qui protègent ce document, Ni la thèse ou le mémoire, ni des extraits substantiels de ce document, ne doivent être imprimés ou autrement reproduits sans l’autorisation de l’auteur.
Afin de se conformer à la Loi canadienne sur la protection des renseignements personnels, quelques formulaires secondaires, coordonnées ou signatures intégrées au texte ont pu être enlevés de ce document. Bien que cela ait pu affecter la pagination, il n’y a aucun contenu manquant.
NOTICE
The author of this thesis or dissertation has granted a nonexciusive license allowing Université de Montréat to reproduce and publish the document, in part or in whole, and in any format, solely for noncommercial educational and research purposes.
Ihe author and co-authors if applicable retain copyright ownership and moral tights in this document. Neither the whole thesis or dissertation, flot substantial extracts from it, may be printed or otherwise reproduced without the author’s permission.
In compliance with the Canadian Privacy Act some supporting forms, contact information or signatures may have been removed from the document. While this may affect the document page count, it does flot represent any Ioss of content from the document. Université de Montréal
Faculté des études supérieures
Cette thèse intitulée
Long-Term Economic Decline and Non-Mainstream Voting
présentée par
Andrea ML. Perrella
a été évaluée par un jury composé des personnes suivantes
Richard Nadeau, président-rapporteur
André Biais, directeur de recherche
Louis Massicotte, membre du jury
François Pétry, examinateur externe
Claire Durand, représentant du doyen de la FES
9 SuMMARY
Canadian voting behaviour from 1974 to 2000 is examined by relating long-term economic changes to support for “non-mainstream” parties, defined as parties other than the Liberals or Progressive
Conservatives. This long-term perspective is unique, in that standard economic voting research focuses on how short-term economic changes affect support for the incumbent. Instead, a long-term perspective shows long-term economic decline yielding a distinct effect. Instead of the incumbent party being evaluated as to its competence over economic management, those who have suffered long-term economic
decline vote in ways that reflect increased rejection for the politicaÏ system as a whole, with this attitude manifesting itself as voter support for non-mainstream parties. This approach, although new in economic voting, is based on established theory as outlined by Lipset (1959/63) and Easton (1975), whereby long-term economic decline is prone to lead voters to acquire attitudes and behaviours that express discontent towards the political system
The impact of long-term economic decline on non-mainstream voting is examined through three different approaches, presented as three separate articles. Each, in its own way, illustrates the effects of long-term economic decline. The first article employs an aggregate approacli. Federal voting resuits from 1979 to 2000 are related to short- and long-term economic data, namely unemployment and
3 labour-force participation rates, ail aggregated at the provincial level.
The pooled data produce resuits that confirm the relevance of short term changes to explain support for the incumbent party, as hypothesized by the “responsibility hypothesis” of standard economic voting, while support for non-mainstream parties is explained by long term economic decline.
The second article examines long-term economic decline through the perspective of occupation. Data from the 1979 and 1993 Canadian
Election Studies and the 1971, 1981 and 1993 Canada Census are pooled together for an individual-level analysis. Resuits show that those whose occupations have suffered long-term economic decline feel
Ïess external political efficacy, vote less, and are more hkely to support non-mainstream parties.
The third article draws upon some of the findings from Teixeira and Rogers’s study (2000). Canada is used as a case study to demonstrate how the long-term economic decline of working-class individuals who do not possess post-secondary education has changed this group’s voting behaviour in a manner that reflects withdrawal of support for the political system as a whole. As in Article 2, both objective economic data (Statistics Canada) and election surveys (CES
1974 to 1997) are used to build models that show how members of this disadvantaged group behaved no differently from other voters before the 1980s, when economic conditions were more “equitable,” and liow after the early 1980s their long-term economic decline eroded further
4 O theii- sense of external political efficacv, which in turn lcd them to vote increasinglv for non-mainstream parties.
The project concludes bv highlighting two general flndings. First,
there is a qualitative difference between short-term and long-term
economic experience. The former yields implications about the
competence of the incumbent, while the latter sheds light on support
for the entire political system. Using the multi-stage vote model
developed by Blais et al. (2002) as a tool for discussion, economics is
highuighted as a central component of the vote calculus, and a factor
that yields implications both to more proximate factors as well as to
established attitudes and values.
The second general finding concerns the consistencv of the
resuits. Generally, the link between long-term economic decline and
support for non-mainstream parties is found no matter what
perspective is utilized. However, other factors, such as the region in
which a voter lives, appear to bear more weight. Nonetheless, given the
fact that long-term economic decline affects a voter’s attitudes towards
the entire political system, and given the fact that it appears to take a
long time to develop these attitudes, these effects may be durable.
The conclusion also points to some potential objections \vith the
key concepts of the project and the methods that were employed.
Key words: long-term economic decline; non-mainstream parties;
political efficacy; economic voting.
5 RÉSUMÉ
Le comportement électoral des Canadiens entre 1974 à 2000 est examiné en reliant les changements économiques à long terme et l’appui accordé aux partis « non dominants », à savoir les partis autrc que le Parti libéral et le Parti conservateur. Cette perspective à long terme est unique en son genre car les recherches courantes sur le vote
économique étudient surtout la relation entre le ressentiment des
électeurs, les changements économiques à court terme et l’appui au parti sortant. À l’inverse, dans cette thèse, une perspective à long terme est adoptée et les résultats démontrent que le déclin économique à long terme entraine un effet distinct. Au lieu d’évaluer la gestion
économique du parti sortant, les individus qui ont subi un déclin
économique à long terme adoptent des comportements politiques suggérant un rejet du système politique en général, lequel se manifeste clairement par l’appui à des partis « non dominants ». Cette approche, bien que nouvelle dans le domaine de vote économique, est basée sur la théorie déjà établie et décrite par Lipset (1959/63) et Easton (1975), selon laquelle le déclin économique à long terme mène les électeurs à adopter des attitudes et des comportements qui suggèrent le mécontentement envers le système politique.
L’impact du déclin économique à long terme sur le vote est examiné par trois approches différentes, présentées en dans trois articles séparés. Chacun, de sa propre manière, illustre les effets du
6 déclin économique à long terme. Le premier article utilise une approche agrégée. Les résultats des élections fédérales de 1979 à 2000 sont reliés à des données économiques à court et à long terme, notamment les taux de chômage et d’occupation de la population active, calculés à l’échelon provincial. Les données agrégées donnent des résultats qui confirment la pertinence des changements à court terme pour expliquer l’appui au parti sortant, tandis que le soutien aux partis « non dominants » s’explique au contraire par les changements
économiques à long terme.
Le deuxième article examine le déclin économique à long terme par la perspective du statut professionnel. Les données des différentes
études électorales canadiennes de 1979 à 1993 et les données du recensement canadien de 1971, 1981 et 1993 sont mises utilisées pour poursuive une analyse au niveau individuel. Selon les résultats obtenus, les individus pratiquant un métier marqué par le déclin
économique à long terme tendent à percevoir moins d’efficacité
politique des gouvernements, à voter moins, et à appuyer des partis
non dominants ».
Le troisième article est basé sur l’approche adoptée par Teixeira et Rogers (2000), et appliquée aux tats-Unis. En examinant la situation d’un groupe démographique canadien particulier, soit les membres de la classe ouvrière ne possédant pas un diplôme d’éducation postsecondaire, cet article démontre comment le déclin
économique à long terme a changé leur comportement électoral et
7 mene au retrait de leur soutien au système politique dans son ensemble. Cet article reprend également des données économiques
(Statistique Canada) et des enquêtes électorales (EEC 1974 à 1997).
D’une part, les résultats montrent clairement que les membres de ce groupe, déjà désavantagé, se sont comportés comme les autres
électeurs avant les années 1980, soit à l’époque où les conditions
économiques étaient plus « équitables «. D’autre part, ils démontrent
également que, depuis le début des années 1980, leur déclin
économique à long terme a mené à une augmentation de leur mécontentement à l’égard du système politique, ce qui a causé l’augmentation du vote pour des partis non dominants.
La thèse se termine en soulignant detix résultats généraux.
Dabord, il y a une différence qualitative entre les effets économiques à court et à longs termes. La perception des effets économiques à court terme a des implications sur l’opinion à l’égard de la compétence du parti sortant, alors que celle à long terme a un impact sur le système politique dans son ensemble. En utilisant le modèle de vote « multi niveaux » développé par Blais et al. (2002) comme outil pour la discussion, les conditions économiques sont considérées comme un facteur central dans les choix électoraux. Plus précisément, elles semblent avoir un effet sur les éléments les plus proche, mais
égah ment un effet sur les attitudes et valeurs déjà établies.
Ensuite, la deuxième conclusion générale concerne l’uniformité des résultats. D’une façon générale, le lien entre le déclin économique
$ à long terme et l’appui de partis non dominants est trouvé peù importe la perspective analytique adoptée. Cependant, d autres facteurs, tels que la région dans laquelle un électeur réside, semblent avoir plus de poids. Néanmoins, étant donné le fait que le déclin économique à long terme affecte les comportements dun électeur envers le système politique pris dans son ensemble, et étant donné que cela a pris beaucoup de temps pour que ces attitudes se développent, ces effets ne sont pas susceptibles de disparaitre rapidement.
La conclusion souligne aussi quelques objections potentielles avec les concepts principaux du projet et des méthodes qui ont été utilisés.
Mots clés t déclin économique à long terme; partis « non dominants » efficacité politique; vote économique.
9 TABLE 0F CONTENTS
Summarv 3
Résumé 6
ListofTables 13
Listoffigures . . . 14
Acknowledgements 15
Introduction: Economic foundations of Political Legitimacy 17 A- Current Research: The Short-Term Perspective 23 1- Responsibility Hypothesis 24 2- Policy Oriented Voting 28 B- Operationalizing Voter Behaviour 31 C- Methodology and Hypotheses 37 D- General findings 45
Chapter 1: Long-Term Economic Hardship and Non-Mainstream Voting in Canada 47 Introduction Economic Decline and Voter Volatility 48 Conventional Economic Voting Models 53 Conceptualizing Voting: Mainstream vs. Non-Mainstream 59
10 Data and Methods . 66
Resuits . 72 Conclusion 79
Chapter 2: Long-Term Economic Decline Among Occupational Groups: Political Efficacy and Support for Non-Mainstream Parties $4 Democratic Choices: Turnover or Rage 85 Narrow Focus of Standard Economic Voting Research $9 Non-Mainstream Parties 92 External Efficacy: Measure of System Support 94 Data and Methods 97 Resuits 103 Conclusion 111
Chapter 3: Economic Disparities and Non-Mainstream Voting: The Discontent of Working-Class Voters with No Post-Secondary Education 114 Introduction 115 Re-Conceptualizing Economic Voting 122
Diverging Economic Demographic Groups . . . 128 Data and Methods 132 Conclusion 146
Conclusion 149
1- Long-Term vs. Short-Term: A Qualitative Distinction. . . 152 2- Resiliency of Resuits 157 3- Potential Objections and Qualifications 164 Q 4- final Remarks 176
11 Q BibÏiograpliy . 180
Appendix 1 Outiine of Variables for Article 1 . 19 1
Appendix 2: Outiine of Variables for Article 2 195
Appendix 3: Outiine of Variables for Article 3 19$
o
12 O LIST 0F TABLES
Table I. 1 Basic summary of research project 38
Table 1. 1 Non-mainstream parties that ran candidates in elections from 1979 to 2000 65
Table 1.2 Comparing long— and short-term effects 73
Table 2. 1 Job growth for occupational groups, 1971-1991 98
Table 2.2 Multinomial logit estimates of vote choice, 1979 & 1993 Canadian elections 104
Table 2.3 OLS regression of political efficacv, 1979 & 1993 Canadian elections 108
Table 2.4 Multinomial logit estimates of vote choice, 1979 & 1993 Canadian elections 110
Table 3.la Multinomial logit estimates of vote choice, 1984 to 1997 Canadian elections 136
Table 3.lb Multinomiallogit estimates of vote choice, 1974 to 1980 Canadian elections 137
Table 3.2 OLS regression analysis of external political efficacy, 1974 to 1997 Canadian elections 144
Table 3.3 Multinomial logit estimates cf vote choice, 1984 to 1997 Canadian elections 145
Table C. 1 Factors that influence non-mainstream voting 158
Table C.2 factors affecting political efficacy, comparing beta scores from Articles 2 and 3 162
13 LIST 0F FIGuis
Figure 1. 1 Vote shares of mainstream and non-mainstream parties 66
Figure 1.2 Labour market trends in Canada, 1969-2000 69
Figure 1.3a Participation rate and non-mainstream vote, Western provinces 74
Figure 1.3b Participation rate and non-mainstream vote, Central Canada 74 figure 13e Participation rate and non-mainstream vote, Atiantic provinces 75
Figure 2. 1 Simulation of vote probabilities and occupational-level economic change 105
Figure 3.1 Unemployment rates 130 figure 3.2 Political inefficacy, 1974-1997 141
Figure 3.3 Simulation of non-mainstream voting and political inefficacy 146
Figure C. 1 Multi-stage vote model 153
Figure C.2 Modffied multi-stage vote model 156
14 AcKN0wLEDGEMENTs
The completion of this project could not have been possible
without the thorough and meticulous guidance from my research
director, Prof. André Biais. From beginning to end, Dr. Biais performed
his duties as a supervisor in a way that made me often feel that he
functioned as a partner in this endeavour. Our discussions helped
bring greater precision to my work, and bis experienced council helped
me grasp the ropes of academic life. A graduate student could not be in
better hands.
Others have also lent an invaluable hand. in particular, Prof.
Richard Nadeau, who was instrumental at the early stage of this
researcli, and Prof. Robert Dalpé, whose seminar on thesis-writing
helped get the bail rolling.
Most of ail, I would like to thank my family. I am gratefui to my
parents who sacrificed a great deal to see me get an education. And I
especialiy want to thank my wife, Christine, who provided me with
much-needed support and motivation, especiaily towards the end of
this project. My heartfelt thanks to you ail. o
15 o
“Where sorne peopte are very wealthy artd others
have nothing, the resutt wiÏÏ be either extrerne
dernocracy or absolu te oÏigarchy, or despotisrn
wiÏl corne from either of those excesses,”
— Aristotie
o
16 INTRODUCTION: EcoNoMIc F0uNDATI0Ns 0F
PoLITIc LEGITIMAcY
The link between economic hardship and voting behaviour ‘s examined by paying particular attention to the last 30 years. Since the
1970s, voters have witnessed turbulent economic changes. The
Keynesian economic orthodoxv that sustained generous welfare programs came under attack. Also, the past 30 years has been marked by turbulent election resuits, such as the minority governments of the
1970s and the emergence in the 1990s of new parties that criticize basic fundamental Canadian political institutions, namely the Reform/
Alliance Party, which advocated constitutional change in order to elevate the influence of Western provinces, and the separatist Bloc
Québécois. The basis of support for such parties, arid other, smaller parties that have not acquired as mucli prominence, are examined in light of long-term economic decline.
Economic hardship is often associated with political instability.
That people who experience economic problems will look for new political solutions is an idea that has great intuitive plausibility. One possible outcome of economic hardship may be as little as a change in governing party, such as in 1984 when the Progressive Conservatives defeated the Liberal Party, which governed during the recession of the
1980s. Another, and less frequent, consequence of economic hardship is outright rebellion. In the former case, voters react to the performance
17 of incumbent politicians. In the second case, voters react to a perceived
dysfunction in the entire political system. The latter case holds far more drastic implications for democratic governance. But given the current behaviour of voters in Canada, it is no entirely clear whether
supporters of the new and semi-successful third parties1 are expressing a desire for a simple change in governance, or whether they are reacting against the system as a whole. We know even less about the basis for support of more marginal parties that span the political spectrum, be they the communist parties on the left, the Libertarians on the riglit, or any of several religious parties (e.g., Christian Heritage,
Natural Law, etc.). In any case, a generation of economic hardship may be a common driver among voters who have turned away from both the
Liberals and the PC Party and towards parties that espouse a completely new political agenda.
The reason economics is advanced as a potentially key determinant of voter discontent stems from the extensive literature that links together economic hardship and political volatility. Lipset
(1959/63) notes that one feature of a stable government system is its ability to ensure a certain level of general prosperity, which sustains the system’s legitimacy. “A society divided between a large impoverished mass and a small favoured elite results either in oligarchy (dictatorial rule of the small upper stratum) or in tyranny
1 Both the Reform/Alliance and the Bloc Québécois have von enougli seats to form the officiai opposition.
18 (popu1ar-based dictatorship),” (Lipset, 1959/63: 31). If an individual no
longer feels assured of a certain level of well being, the legitimacy of the
system mav erode. furthermore, in situations in which conditions are
in flux, a new group of winners may emerge at the expense of a new
group of losers, with the losers reacting by supporting movements that
question the legitimacy of the regime, giving rise to new cleavages, or
even making already existing cleavages more poignant.
A well known Canadian case study of the economic basis of voter
discontent is by Maurice Pinard (1971). Following Smelser’s (1963)
tlieory of collective behaviour, Pinard shows that several factors were in
place that favoured the growth of the Créditistes. In particular, Quebec
voters who were suffering economic decÏine (or “strains”) and who also
lived in rural regions made it conducive to mobilize voters around this new party. Such xoters switched from one major partv to another,
eventually giving up on both to support the new political movement.
Both Lipset and Pinard see the link between economic conditions and voting behaviour as reflecting more than just an evaluation of the competence of political leaders. Instead, such a link is
a manifestation of suppurt — or lack, thereof — for the entire political system.
More recent work in economic voting research does not accommodate such an understanding of voting. Lewis-Beck (195$), for instance, confirms a link between economic downturn and declining support for the government, relying on the standard economic voting
19 Ç researcli approach relating election survev questions on economic perceptions and vote intentions. The main focus is on the incumbent
party and its overali level of support, vis-e vis other parties. The main
method is to track short-term changes in voting intentions as
expressed in election surveys, with survev questions that ask
respondents to evaluate the economy, mostly within a short-term time
frame (usually 12 months). Conclusions point to a direct relationship
between general economic decline and general decline in popular
support for the governing party.
This approach may be suitable to forecast election resuits, but
the narrow focus on the short-term is deeply unsatisfactory. Standard
economic voting literature makes no distinction of vote shifts from the
incumbent mainstream party to a mainstream opposition party,
compared to a vote shift from the incumbent to a different type of
opposition party, one that is more ideological, perhaps more radical,
and perhaps more critical of the “state,” and not just critical of some
policies of the incumbent party. This is an important omission in the
literature. In essence, standard economic voting research fails to
distinguish between short-term anger against an incumbent partv and
an outnght legitimacy crisis. This is probably due to reliance on survev
data, where sample size may make it next to impossible to trace
developments of discontent at the fringes of a party system; too few
respondents express support for other parties. Therefore, research that o
20 relies solely on survey data finds it mucli easier to measure support for and against the incumbent partv.
Although the project being presented here also makes use of survey data. the pooling together of several surveys permits the capture of voting at the fringes. In Article 1, actual electoral resuÏts from elections held between 1979 and 2000 for each party are used. Articles
2 and 3 use Canadian Election Survey data from 1974 to 1997. This way, distinctions can be made about shifts in voter support between mainstream “pragmatic” parties and non-mainstream parties.
There has hardly been any analysis on whether a shift away from incumbent parties benefits just another incumbent and pragmatic party, or a party with more radical appeals. In other words, economic voting bas not yet addressed the issue of the potential for long-term economic hardship on voter attitudes about the political system as a whol e.
This project focuses on long-term economic decline, particularlv since the 1970s, and its link to voter support for parties that challenge the status quo. 0f all the major events in Canadian elections since the
1970s, none ranks more prominent than the 1993 general election, an election marked by two main developments: 1) the rise to prominence
of two non-mainstream parties; and 2) the demise of the Progressive
Conservative Party, one of the oldest and most estabÏished mainstream
parties of Canada. This most recent episode of the development of third
parties in Canada may validate Pinard’s theory, such that a generation
21 O of economic decline brought upon bv the stagflationa 1970s 1980s may have moved some segments of the electorate to shift
support from the Liberals and PC to Reform/Alliance and the Bloc
Québécois. Perhaps 20 vears cf economic turbulence, and the forces
put upon governments to curtail expenditures and redistributory
policies, have lcd to a whole new set of winners and losers, which, as
Lipset would suggest, has lcd to the emergence of a an issue dimension
along which voters divide. But you wouldn’t know that from the
standard economic voting literature, which makes no attempt to
measure the long-term influence of economic stagnation.
In sum, this project addresses the general issue of stable
democratic governance bv looking at voting as an indicator of
discontent towards the entire political system, and by looking at long
term economic hardship as a contributing factor of discontent. The
project draws a distinction between voting for a mainstream and non
mainstream party. A switch from one mainstream party to another will
be seen as less significant than a switch from a mainstream party to a
non-mainstream party. Most economic voting literature makes no
distinction between the political signiflcance of choice along these unes.
The project makes the assumption that sometimes voters feel drawn to
choices that would otherwise seem unattractive under more favourable
economic conditions, and that long-term deteriorating economic
conditions can disconnect voters from their traditional mainstream
party choices.
22 A- Current Research: The Short-Term Perspective
The following literature review focuses on two main different approaches to understand the link between economic conditions and voter behaviour. The first section on the “Responsibility Hypothesis” reviews the more mainstream approacli, characterized by a high reliance on election survey data and a focus on short-term changes.
The second section, entitled “Policy Oriented Voting,” focuses on a literature that places more centrality on personal economic conditions and their salience to a voter’s views on liow the state should govern. This stream of research argues that it may take several elections for long-term economic hardship to motivate voters to abandon mainstream parties in favour of newer political movements.
Therefore, the mainstream focus on short-term changes may be unsuitable to capture a phenomenon that spans a longer period of time.
Economic hardship also combines with different factors to become eïectorally salient. Individuals suffering hardship can relate to others in a similar predicament, at the regional, sectoral, and even the class level, leading to a collective consciousness that propels a new movement.
23 1- Responsibilit Htjpothesis
The foundation of economic voting Tests Ofl the so-called
“responsibility hypothesis,” where the competence of the incumbent
government is evaluated in light of how it handled the economy. One of
the first empirical tests of the hypothesis thesis cornes from an analysis
of House of Representatives elections in the United States. Kramer
(1971) ernploys aggregate data on income and overali support for
incumbent party to validate the hypothesis. Drops in income are
inversely related to support for the incumbent party. Kramer’s use of
“income” can be considered an egotropic indicator, but lis aggregate
analysis is consistent with eitlier a sociotropic or an egotropic
conclusion, given that by aggregating income data, it is not clear
whether voters are responding to changes in their personal
circumstances, which would make it egotropic, or to changes in overali
national income levels, a sociotropic phenomenon (Kramer, 1983).
furthermore, it must be pointed out that Kramer’s analysis relates
short-term changes in income to short-term changes in support levels.
An individual-level confirmation of the responsibility hypothesis
was conducted by Fiorina (1978). Using survey data, Fiorina concludes that voters respond to changes in economic conditions retrospectively, whereby past economic conditions weigh in highly. Also, voter reactions appear sensitive to short-term changes. fiorina notes, however, that economics is not aiways politically salient.
24 Kinder and Kiewiet (C79) coauthored a study that further
substantiates the responsibiÏity hypothesis. According to the author’
analysis of a data series stretching from 1956 to 1976, voters appear to
switch party preferences according to changes in national economic
conditions. Their analysis relies on National Election Study survev
data, relating short-term changes in perceptions of economic
conditions to vote choice.
Many other studies reinforced the sociotropic version of the
responsibility hypothesis, with more nuanced findings. For instance, on
the question of which economic indicatnr is most politically salient,
some studies point to unemployment, but not just the raw rate. It may
be important that unemployment is high or low in absolute terms, but
what seems at least equally as important, if not more so, are changes
in the unemployment rate (Nadeau and Blais, 1993, 1995), as well as
perceptions about job growth, especially if expectations exceed what is
reported by government statistics.
Inflation is another key economic indicator. As Chappeil and
Viega point out (2000), inflation is one economic condition that the
government can control through price controls and perhaps even
througli monetary poÏicy. But the relationship between inflation and
support for the incumbent depends also on the particular party in
power. As per the clientele and “salient goal” hypotheses, the relative
salience of individual economic indicators depends on a partv’s general () policy agenda. Some parties “own” a particular issue (Carisen, 2000;
25 Hibbs, 1987; Rattinger, 1991; Swank, 1993). The United States is a
good example where the Democrats foLus on unemployment, whule the
Republicans focus on inflation. 2 Therefore, in times of higli
unemployment, the Democrats are deemed to benefit more in
opposition, because they are seen as the party that is most dedicated to
job growth. However, a rise in unemplovment is seen to hurt a left-of
centre party more if it’s in government compared to its potential to
erode support for a right-of-centre administration (Goodhart and
Bhansali, 1970). furthermore, while in government, voters adjust their
expectations, figuring a left-of-centre government is probablv more
likely to make more progress on unemployment than a more right-of
centre government (Poweil and Whitten, 1993). It should be noted
according to studies of “independents” in the Untied States, sucli
voters rely more on egotropic evaluations, given that such voters are
disconnected from partisan cues on how to evaluate sociotropic
conditions (Romero and Stambough, 1996).
Overali, the responsibility hypothesis sees voters as evaluators of
government performance, relying mainly on national economic
conditions. Personal finances are flot seen to measure highly in the
evaluations. One reason given for this finding is the difficulties in
linking personal finances to government policy. An “ethic of self-
2 American writer Ambrose Bierce (1842-1914) once joked: “What is a Democrat? One who believes that the Republicans have ruined the country. What is a Republican? One who believes that the Democrats would ruin the O country.”
26 reliance” may explam why voters, especially in the individualistic
cuÏtLlre of the United States, do not link personal fortunes to
government behaviour (Sniderman and Brodv, 1977). Therefore, the
government is seen as vielding less control on personal finances.
However, changes to some aspects of personal finances can be traced
to government policy. feidman (1985) notes that although overail
sociotropic voting is more prevalent than egotropic voting, there is a
“mediated pocketbook effect” where personal finances, or a part
thereof, are linked to government behaviour. Peidman also notes that
this otherwise negligible effect could be compounded by general
conditions: “These resuits provide clear evidence that under certain
conditions — poor economic conditions and clear attributions of
responsibihty — personaJ economic self-interest can play a role in
political evaluation,” (p. 159).
Perhaps the strongest evidence for an egotropic effect cornes
from the welfare states of nortliern Europe. In Scandinavian countries,
where the government takes a more active role in individual economic
well being, personal finances appear more politically relevant
(Nannestad and Paldam, 1994, 1995, 1997). Lewis-Beck (1986, 1988)
also found evidence for pocketbook voting, but along with evidence for
sociotropic voting. Using the Euro-Barorneter surveys, Lewis-Beck analyzed economic voting in Western European countries, and developed a model that encompasses different variables that track both general economic trends and more personal, household, conditions.
27 Pocketbook evaluations, however, show negligible effects, overail. Yet.
similar to what was found in the Scandinavian countries, Lewis-Beck found that “when voters think that government policies have had a
good or a bad effect on the household financial situation, their vote
intention is significantly influenced,” (Lewis-Beck, 1986: 325).
Overali, the responsibility hypothesis concludes that voters rely
predominantly on retrospective evaluations of sociotropic conditions in
order to render a verdict on how well the incumbents are governing.
The economy is an important factor, both because it is an important
part of life, and because it represents a symbol of overali governance. If
the overali economy seems prosperous, voters infer that the
government is doing a good job overali. Sociotropic conditions are also widely reported by the media, so voters do not need to possess a great
deal of information in order to conclu de whether or not the economy is
functioning well (Anderson, 2000; Key, 1966). Even thougli personal
finances may provide a more immediate source of economic
information, personal circumstances are seen as related more to
personal efforts (feldman, 1982).
2- Polict Oriented Voting
While the responsibility hypothesis relates short-term self-report
perceptions of economic changes to short-term changes in party
28 preferences, a longer-term view reveals voting behaviour as reflective of policy evaluations (Brooks and Brady, 1999). Voters are flot understood simply to pass judgment on the government’s competence. Instead, voters take a more global perspective; they interpret th legitimacv of changes to welfare programs, cuts to the civil service, particular public policies, and so forth. for example, Brooks and Brady point to the New
Deal as an income redistribution policy that sets the foundation for why certain voters have divided themselves between the Republicans and Democrats. The more redistribution-friendly Democrats tend to draw the support of poorer voters, whule the wealthier tend to back the more business-friendly Republicans. According to Brooks and Brady,
“policy evaluations of the welfare state explain 33% of the differenc’s between voters in the top versus bottom quintile of the income distribution,” (1999: 1361). The attack on the welfare state has created a new underciass with an angrv baggage of attitudes towards any of the major parties, and perhaps towards the system as a whole.
This une of work is echoed by Teixeira and Rogers (2000), who demonstrate how the more affluent period before ihe early 1970s gave way to a growing income gap between the upper and lower classes, a gap that lias led to a level of discontent among what they refer to as the
“forgotten majority,” namely white working-class individuals wlio do not possess a college education (see also Teixeira, 1998). Before 1973, ail hard-working Americans could expect to become middle-class.
Changes in general economic conditions were feit more equitably
29 across the board. But after 1973, a widening wage gap raised the
salience of long-term personal economic conditions among that
segment of the voting population that was most adversely affected. As
Nagler and De Boef (2000) point out, if national conditions improve
while personal conditions do not, then voters look for other indicators,
namely their “economic reference group,” such as occupational peers. If
individuals’ deteriorating personal situations are matched by their
peers, then voters might rely more heavily on this experience when
evaluating how well the government is looking after their irierests. It is
precisely this approach that guided the analysis reported in Article 2.
Pinard (1971) also points to reference groups as important to
integrate voters to a new movement. An individual’s self-identification
is partly attributed to memberships in various primary groups, groups
that often define themselves according to ethnicity, language, religion,
and region. Membership in these groups become politically salient in cases where the group clearly finds itself at a disadvantage, where there appear to be “in groups” and “out groups” in how political benefits are distributed. Lipset (1959/63) discusses this phenomenon in light of economic changes, where changes in stratification lead to the development of new political cleavages. However, Pinard focuses on primary groups as an especiaiïy salient factor to explain why Social
Credit succeeded more in rural areas f Quebec, where voters are more integrated into primary groups, whule the more autonomous urban voters were less likely to be integrated into a new movement. A highly
30 integrated electorate recognizes peers who share a similar negative
economic experience, and this may lead them to a rejection of the
current siate of mainstream parties as inadequate.
Another interesting feature of understanding voters through the
perspective of referene groups is its challenge of the commonly
asserted notion in standard economic voting research of voter myopia.
Voters are understood to have short-term memories. However, as noted
by Van Der Brug et al. (2000), voters’ memories may lie organized more
intricatelv if understood in terms of a voter’s membership to a reference
group. It may not necessarily lie true that voters simply follow their
herd, but it seems reasonable to expect personal concerns, such as
hardship, to become politically salient through a reference group.
Standard economic voting researcli lias not found sucli a phenomenon,
but this seems most likelv due to the fact that it lias not looked for it.
B- Operationalizing Voter Behavïour
Several hypotheses are tested to determine whether long-term
economic hardship have an effect on the vote. The’ will be outlined in
detail shortly. However, before going into further detail, it should be
made clear what is meant by “voting behaviour.” Voter behaviour here
is examined in ways that lead to an interpretation of voter support for the political system as a whole. A developing legitimacy crisis among
31 voters, even just a small segment of voters, has not been studied
carefully enough. How voters express such sentiments is also net clearly established. But it seems plausible that voters upset with the entire political system would be more prone to abandon those parties most closely associated with the political system. Here, this is viewed as a vote for a party other than that for the Progressive Conservative or the Liberal parties. A vote for any other party is a vote for a non mainstream party.
No distinction will be made within this broad category.
Surprisingly, this is a unique approach since most studies of sucli parties tend to focus on particular sub-types of parties, especially those that are most prominent. For example. in Canada, Pinard’s (1971,
1973) work on the rise of “third parties” was focused on Social Credit.
Lately, a considerable amount of attention has been given to the phenomenon of anti-state and anti-system voting, phenomena that appears to have grown more prominent over the last few decades in many advanced democratic states (Bélanger, 2004; Clarke and
Kornberg, 1993, 1996; Clarke, et al., 2000; Gidengil, et al., 2001; see also Poguntke, 1996, and Sartori, 1976). More recent work focuses on how discontented voters would be more prone to abandon both the incumbent and the mainstream opposition party if there is a viable
“third party” choice (see Hetherington, 1999). Some other work shows a shift of non-mainstream voting from Canada’s “traditional” third party,
32 the NDP, to the newer upstarts Reform/Alliancc and Bloc Québécois
(Belanger and Nadeau, 2005).
But what is missing in this type of resear h is a clear distinction between mainstream and non-mainstream. Although the term
“mainstream party” bas appeared in the literature (in fact, it bas appeared quite a lot), it has not aiways been adequately defmed. for example, a report of the 2003 Scottish elections (Burnside et al., 2003) refers to the top four parties as mainstream. Meguid (2005) provides a more precise definition: “Mainstream parties are defined as the electorally dominant actors in the center-left, center, and center-right blocs on the Left-Right political spectrum,” (352). A problem with this definition is that it connects mainstream to the ideological spectrum.
That may be fine when one describes an extreme left or extreme right party as non-mainstream, but according to the logic, the doser a party is to the center, the more it qualifies as mainstream. But one must not equate a rnoderate party with the mainstream. There ai-e centrists parties that fail to attain the respectability afforded to a so-called mainstream party. In other words, it may be necessary for a mainstream party to hover the center, but a centrist party does flot necessarily qualify as mainstream.
An even less clear concept is “non-mainstream.” Ibis term has not received nearly as many mentions, and is even less likely to be described in any precise manner. At most, non-mainstream implies the fringe or extreme end of a party system, i.e., the leftovers of a party
33 (N system. For example, Meguid (2005) distinguishes mainstream parties
from “niche” parties, which tend to focus on single issues or a narrow
segment of the electorate. These include Green parties, which tend to
focus primarily on the environmentai issue, and some radical-right
parties, which tend to assert a platform focused on immigration and
welfare reform, and not much else. Heari et al. (1996) and Jolly (2004)
distinguish mainstream parties from regional parties. The problem with
these perspectives is that it leaves undefined parties that are not
strictly regional, not necessarily extremist, yet promote views that are
more national and more broad. A non-mainstream party need not
necessarily be either a niche or a regional party. The only thing that
disqualifies these niche or regional parties from the mainstream is their
size: small, marginal parties that win few (if any) seats, or for some
other yet unspecified reason do not qualify as mainstream. As a result,
niche or regional parties tend to do uo better than sit in opposition.
In order to bring more clarity and precision to this concept, a
“mainstream party” is seen here to mean a party that is implicated in
the political system, a political “insider,” even if that party sits in
opposition. In pre-2004 Canada, the Liberal and Progressive
Conservative parties are, indeed, implicated as political insiders, even
when one sits in opposition while the other occupies government
(awaiting its inevitable return to occupy 24 Sussex Drive). Whereas
“non-mainstream” describes a party that is generally regarded as an
outsider, not tainted in any way as having a hand in the political
34 system. In Canada, these inciude ail parties that have not been in power
Arguabiv, this conceptualization may have limited apphcabihty in multi-party systems where unes between “insider” and “outsider” parties may be biurred. Nonetheiess, there is some support in the literature for this definition, although no one bas really gone so far as provide as precise a definition. There is some suggestion that to become more mainstream, a party must be doser to occupying power. There are several ways this can happen. First, a mainstream party can draw on policy ideas of a non-mainstream party, thereby legitimating and providing a measure of respect for a party that was previously seen as non-mainstream (Bale, 2003). Second, a non-mainstream party can be seen as more main stream if it becomes coalitionable, such as the Green parties (Mair, 2001). In either case, what was once an outsider lias
evolved — or is evolving — towards becoming an insider. What was once regarded as a partv that was compietely outside of the political executive is less of an outsider.
Generally, the impetus for such research is the energence of new parties that appear to exploit discontent towards the politicai system and that enjoy some measure of success, even if only at the regional level. The populist movement that spawned the Reform Party in Canada3 and the separatist Bloc Québécois have ail stimulated
The Reform/Alliance populist movement is flot a uniqueiv Canadian phenomenon. In fact, the rise of the radical right in Europe and in other
35 renewed interest in the political basis of these new parties. It is almost as if the growth and “arrivai” of non-mainstream parties is greeted with a sense of surprise. Bu if one looks at election ballots prior to the emergence of any of these “third” parties, one sees parties other than the most prominent. Support for these non-mainstream parties does exist, and support levels do vary from one election to the next. The question then changes from “what caused these parties to emerge?” to
“what explains their unusually high levels of support?” An answer to that question is being pursued here by looking at the economic-basis of support for these parties, even when these parties are not prominent and not a potentialïy “threatening” force.
That is the main purpose that motivated this research project.
Support for non-mainstream parties is the main focus, and thus, support for non-mainstream parties is the main dependent variable.
Support for the incumbent is also studied, but only to show that standard economic voting is correct in that short-term changes to economic conditions yield consequences for the ruling incumbent, but
that it is unabie to explain voter propensity to support non-mainstream
parties.
Attitude measures, such as external political efficacy, have also been used, particularly in the second and third articles. These attitude
advanced industrial countries has drawn a great deal attention. See, for example, Ignazi (2003), the October 2004 issue of Journal ofPotiticaÏ Ideotogies (especiallv Betz and Johnson (2004), Taggart (2004), and Fieschi and Heywood (2004), and Norris (2005).
36 measures validate tlie overail argument that there is an association
between the degree to which voters hold negative feelings about the
political system and the extent to which they support parties that are
clearly not mainstream.
C- Methodology and Hypotheses
Support for non-mainstream parties is related to voters’
experience with long-term economic hardship, as per three different
contexts: region, occupation, and socio-demographic group.4 Each of
these contexts is examined in a unique article. Each article uses
different data, different levels of analysis, and different methods. Table
I. Y provides a brief overview of each.
Economic data are aiways drawn from government statistical
databases, either the Labour Force Survey or the Census. Data were
drawn from 1969 to 2001, although each article focuses on particular
periods. Voting data are drawn from two main sources: Elections
Canada and the Canadian Election Study surveys. The aggregate-level
approach of Article 1 makes use of actual vote resuits as reported by
The concept “class” is being avoided mtentionallv. The term tends to signifv hierarchy and status as it pertains to power relations (i.e., Marxist analvsis), while the purpose here is to focus on a group of voters who share demographic similarities. They mav ah belong to the “working-class,” but only those in the working-class who do not possess post-secondarv education are considered Q members of this economicallv disadvantaged demographic group.
37 ______
0 Elections Canada, whule the individual-level ana1vss in Articles 2 and 3
reÏy on the CES surveys.
Table 1.1 : Basic summary of research project
Article 1 2 3 Unit of Demographic • Region Occupation analysis group Labour Force Labour Force Census (1971- Economic data Survey (1969- I Survey (1976-
- 2000) 1997) - -
-
• Unemployment Economic Proportion of Unempiovment and labour-force
. . indicators . workforce rates participation rates
Canadian Election Canadian Election Elections Canada Voting data Studies (1979 and Studies (1974- 1993) 1997)
Voting Reported vote Reported vote Votes cast rndicators choice choice
Level of Aggregate: . Individual Individual analysis Provinces
Method OLS Regression Multinomial logit Multinomiai logit
In ail cases, a voter’s “long term” economic experience is
understood to mean a span of 10 years. This is regarded as a suitable
period given that it is more than twice as long as the standard
government term, during which voters may have witnessed different Q
38 economic cycles, different election campaigns, and different governing parties.
A series of hypotheses are formulated to explore the relationship between long-term economic hardship and voting intentions. The following three sections give a more detailed description of each article.
ArticÏe 1. RegionaÏ Level
Regional-level analysis is fairly straight forward. Changes in regional economic conditions, as revealed by census data, are related to changes in regional voting patterns. Chapter 1 covers in more detail the rationale behind each indicator, but for now it is sufficient to say that the regional unit of analysis is province, and labour-market indicators are used to measure economic conditions. In particular, short-term and iong-term changes to unemployment and labour-force participation rates are used as key variables to explain voting behaviour, ail at the provincial level. Short-term changes in unemployment are seen as more relevant in explaining voter support for the incumbent, while long-term changes in participation rates, which more adequately indicate the overail economic conditions, reftect more generalized economïc health, are seen as more relevant to explain non-mainstream voting. Election returns from 1979 to 2000 are assembled into one pooled dataset. The provincial-level percentage of vote shares won by the incumbent and by
39 non-mainstream parties (two separate dependent variables) are related to the different economic indicators. These different measures and variables have been used o test the following hypotheses:
H1.1: Support for the incumbent is inversely related to
short-term changes in u nemp toyment.
H1.2: Support for non-mainstream parties is inversety
retated to long-term changes in the labour participation
rate.
Article 2: OccupationaÏ Leuel
Analysis at the occupational level relies on Statistics Canada datasets as well as Canadian Election Study (CES) datasets. The CES tracks respondent occupations, which enables an analysis of the link between the economic conditions of different occupational groups and the vote. Census data are used to identify occupations that have experienced declines. Census data from 1971, 1981 and 1991 are used to track changes over two 1O-year periods: 1971 to 1981, and 1981 to
1991. Occupational-level economic conditions from the first period are related to respondents’ vote choice for the 1979 election, while the
40 second period is related to respondents vote choice of the 1993 CES. A multinomial logit model is generated to test the following hypotheses:
H2.1: Voters in occupationat groups experiencing tower
tevets of job growth are tess tikety to vote and more tikety
to support non-mainstream parties.
The use of CES surveys also allows the inclusion of attitude measures, namely external political efficacy. This attitude was seÏected to determine if it functions as a possible mediating factor between economic decline and vote choice. The logic being that a respondent whose occupation lias declined over the long term is more likely to feel ripped off by the system, and thus, would be more likely to manifest a weaker sense of political efficacy. This attitude, in turn, is expected to raise a voter’s propensity to vote for parties least aligned with the political system, i.e., non-mainstream parties. The potential for such a dynamic is tested with the following hypotheses:
H2.2: Voters in occupationat groups experiencing tower
tevets of job growth regard the potiticat system as less
responsive.
41 H2.3: The relationship between occupational job growth
and voting behaviour disappears when externat politicat
efficacy is inctuded as a control variable.
Article 3: Socio-Demographic Group
Canadian Election Study respondents are categorized according to a socio-demographic group that lias been defined by Teixeira and
Rogers (2000). According to Teixeira and Rogers, non-college educated working-ciass white voters in the United States have experienced long term economic deciine in terms of both income and employment security, ail the whule watching governing elites focus their attention on affirmative action programs and welfare initiatives to help minority groups and other “special interests.” The disadvantaged socio demographic group has consequentiy feit negiected by the system, and after a whuie lias begun to express discontent by supporting less mainstream movements, which in the United States came in the form of Ross Perot’s independent presidentiai bid, or the Repubuican’s 1994 mid-term election campaign which focused on anti-welfare policy positions.
The third article explores whether the same phenomenon lias occurred in Canada. Labour force Survey data have been used to determine whether similar economic trends have occurred in Canada.
42 Using unemploymerit rates, it will be sho\n that those with a working ciass occupation (e.g., traditional “hue collar” trades, factorv workers clerks, etc.) and who lack a coilege education have generaliv been worse off economicaily, but since the 1980s, their precarious economic condition lias worsened re1atve to ail others, and continued to iag weli into the late 1990s. Race was flot examined, because it is flot reported in the Labour force Suruey.
The foilowing hypothesis is tested to determine if the post- 1980s iong-term economic dechne of this demographic group is iinked to increased support for non-mainstream parties.
H3. 1: Working-ctass voters who lack post-secondary
education are more Zikety to vote non-mainstream
compared to ati other voters in etections hetd afler the
1980s compared to elections prior to the 1980s.
CES survey data are used. Respondents who belong to the identified demographic group (working-class, non-coilege educated) are compared to ail other respondents. Datasets from 1974 to 1997 have been pooled into two separate files: one for elections from 1974 to
1980, and the second one for elections from 1984 to 1997.
Similar to the second article, politicai efficacv is examined as a possible mediating factor, and tested according to the foliowing hypotheses:
43 o
H3.2: Working-ctass voters without post-secondary
education have tower levels of externat potiticat efftcacy
than other respondents.
H3.3: Poftticat efficacy is an intervening variable that
exp Zains why working-class voters without post-secondarg
education are more tikely to vote non-mainstream afler the
1980s.
Control Variables
The aforementioned hypotheses may suggest certain conclusions
about the relationship between economic hardship and voting
behaviour, but the apparent link may in fact 5e spurious, or at the very
least may need to be speciEed. The descriptions of Articles 2 and 3
explicitly state political efficacy as a potential mediating factor, but ail
articles, indu ding Article 1, also inciude other control variables.
In ail cases, regional dummy variables are used. Respondents
(or, in the case of Article 1, data points) are identifled as belonging to
the Atlantic provinces, Quebec, or the West. In ail cases, Ontario is set
as the reference group. In the second and third articles, other control
variables have been included, namely: gender (male, not male) religion
44 (Catholic, flot Catholic), language (French. flot french), and age. Also, dummy variables have heen created to account for particular election vears, but onlv for Articles 2 and 3. Election-year dummies were tried in Article 1, but failed to ield signfficant resuits.
D- General Findings
Each article produces findings that confirm the hypotheses. In the first article, nonmainstream voting is explained by long-term economic decline, with short-term decline yielding no impact whatsoever. Converselv, support for the incumbent party is explained by the short-term indicator, flot the long-term measure.
In the second article, survey respondents whose occupation has suffered long-term economic decline show a higlier propensitv to support non—mainstream parties. Furthermore, this behaviour is connected to political efficacy, such that those who work in occupations in decline are more likely to feel less efflcacious.
Results of the third article show that survey respondents who belong to a particular economically hard-hit demographic group
(working-class individuals who do not have post—secondary education) are more likely to vote non-mainstream after 1980, when this group’s economic decline was more pronounced. As in Article 2, this phenomenon is tied to political efficacy.
45 Taken as a whole, these different approaches lead to three
general conclusions. First, long-term economic decline is a relevant
factor. It is able to explain non-mainstream voting scores significant
resuits, even when controlling for other factors.
Second, long-term economic decline yieÏds a quaÏitativeÏv
different type of voting behaviour. Whule short-term decline is more
connected to the evaluation of the competence of particular political
leaders and parties, long-term decline appears more connected to voter
attitudes about the political system as a whole.
Third, it must be pointed out that the effect of long-term
economic decline is flot dramatic. Some of the other control variables
show a greater weight in explaining voter support for non-mainstream
parties, particularly regional factors. The Conclusion discusses this
point at greater length.
o
46 o
CHAPTER 1
LONG-TERM EcoNoMIc HARDsHJP AND
N0N-MAINsTREAM VOTING IN CANADA
This article is published in the Canadian Journal of PoÏiticaÏ Science 0 (2005), 38: 335-357.
47 Introduction: Economic Decline and Voter Volatility
Discontent in Canada rnanifests itself through the ernergence of new parties that erode the support of otherwise well-established parties.
The pluraiity electoral system, which heavily favours few, large and generaily centrist parties, makes such a phenomenon more poignant. In
Canada, the Liberal and Progressive Conservative parties have
traditionally alternated as governrnent and officiai opposition. In such a system, success from any other party is seen as unusual, linked to the rejection of both established parties, and therefore possibly linked to some serious level of discontent. But apart from the rare and often short—lived breakthroughs from “third” parties, Canadians do vote for many other alternatives. When support for such alternatives increase, one cannot but beheve voters are deeply upset about sornething, compelhng them to turn their backs on “politics as usual.”
0f ail the sources of discontent that can dispiace a voter away from any of the two main parties, econornic hardship seems the most obvious. Many other political issues corne and go, but over time economic performance remains a salient consideration among voters. No other issue touches both a voter’s need to survive and a voter’s hopes to prosper. This is why economic conditions are probably the most studied factor in electoral research, and why economic voting researcli offers some of the most robust rnodels that accurateiy predict eiection
outc ornes.
48 Despite its success, economic voting research focuses too
narrowly on support levels for incumbents, which for the most part tend
to be well-established and mainstream parties, by relating such
mainstream support to short-term economic conditions. More often than
not, declines in support for the incumbent tend to correspond to
increased support for another mainstream opposition party.
Traditionally in Canada, this support tends to volley between the Liberal
and Progressive Conservative Parties. In the United States, the
Republican and Democratic Parties. For the most part, such movements
guide the headiines are voters supporting the current government or
the main opposition?
Although a short-term approach helps explain why voters like or
dislike an incumbent, it does not adequately explain the strengthening
or weakening support for alternative parties, many of which advance a
radical agenda. This neglect stems largely from the focus on short-term
economic and political changes, an approach that makes it difficuit to
study the basis of electoral support for smaller parties, the support of
which may depend more on structural, longer-term conditions. But a
look at electoral politics in Canada taud other countries) over the past
generation suggests a need to take a more complete look at voting
behaviour in light of economic changes. Many reasons might explain the
emergence of new, and one might argue, radical, parties since 1993,
namely the Reform/Alliance party, which advocates constitutional
(3 change in order to elevate the influence of western provinces, and the
49 separatist Bloc Québécois, which advocates a different sort of
constitutional change. Can it not be argued that suppc rt for these
alternative parties stems from a deep level of discontent, which might be
tied to long-term economic decline?
Such a question is a contrast to the concerns of conventional
economic voting research. A recession here and there is not unusual
and not expected to lead to the same level of rage as when voters
persistently find themselves struggiing economically, even between
recessions. Such a discouraging experience might lead one to question
not only the sitting incumbent, but might lead to the development of
more severe evaluations that pin blame on the political structure, given
that it is seen to fail in delivering the goods regardless who’s in power.
In a way, there is nothing new to the notion that voter discontent
could stem from economic hardship. Lipset (1959/1963) states, for
instance, that one feature of a stable government system is its ability to
sustain a healthy economy, which sustains the system’s legitimacv.
“From Aristotle down to the present, men have argued that only in a
wealthy society in which relatively few citizens lived at the level of real
poverty could there be a situation in which the mass of the population
intelligently participate in politics and develop the seif-restraint
necessary to avoid succumbing to the appeals of irresponsible
demagogues,” (30). Whether or not non-mainstream parties are led by
“irresponsible demagogues” is a separate issue, but Lipset’s main point O is that voters faced with a serious economic decline tend to abandon the
50 “usual” parties in favour of something completelv new. Furthermore,
Lipset notes that this is especially true among those who experience a
great deal of economic volatility, such as miners, farmers, and so forth.
He adds that discontented people who share a common bond or
common communication channel are more likelv to mobilize politically.
This helps explain wliv manv new pohtical movements tend to succeed
flrst in particular regions (e.g., agricultural regions) or among particular
segments of a population (e.g., unionized workers).
In Canada, the political implications of economic hardship have
been explained most thoroughly by Maurice Pinard (1971), who focused
on how economic strains helped shift Quebec voters awav from the
Liberals and Progressive Conservatives and towards Social Credit. He
demonstrated how voters in economicallv hard-hit regions are more
likely to turn to a third partv for solutions, once voters reject the
traditional parties for doing too littie. Pinard used his findings to
validate Smesler’s (1963) theory of collective behaviour, whereby strains
(such as economic hardship), combined with other conditions, such as
the ease with which individuals can mobilize, lead to unrest. Sucli
conditions were present in rural Quebec, where voters suffering
economic strains switched from one major party to another, eventuallv
giving up on both to support a new political movement. It should be
pointed out that according to Pinard, it is not poverty per se that triggers
discontent, since the poor are often disengaged from politics. But a C
51 change from one economic condition o a worse one can lead to support
for a new political movement, independent of the movement’s ideology. I
A common feature of Lipset’s and Pinaid’s perspectives is the
attention paid to voting as flot just an expression of approval or
disapproval for the governing party, but also as an expression of support
for the entire political system. This is measured by the extent to which
economic decline places voters in an increasmgly \ulnerable position,
and consequently, leads them to question the appropriateness, or
legitimacy, of their regime. Legitimacy and stability, curiously, are two
central concerns in political science that contemporary work in
economic voting appears to overlook. Perhaps Canada is not a likely
candidate for an ali-out legitimacy crisis, but the research here will
show that Canada is also not immune to such a development. Long-term
economic decline is related to reduced support for mainstream political
parties, but such a development escapes conventional economic voting
models, hence the need to re-conceptualize the link between economic
conditions and voting behaviour.
I Pinard (1971) notes that the success of a new political movement is flot as dependent on its ideology, given that discontented voters who are readv for an alternative would tend simplv to support whichever partv “that appears to them most likely to be successfuÏ, whether it is conservative or progressive in the eves of tlie sophisticated observer,” (95), which was the Social Credit in the case Q of Quebec.
52 Conventional Economic Voting Models
A great deal cf economic voting research is based on the
“responsibility hypothesis,” where voters are understood to pass
judgment on how well the government has handled the economy. This is
by far one cf the most proiffic areas of voting researcli, having
accumulated a wealtli of literature that includes hundreds cf
manuscripts (Lewis-Beck and Paldam, 2000). It is no surprise, since in a
way, it is rather easy to formulate testable liypotheses regarding
economic conditions and voting behaviour. Models that predict election
prospects for the incumbent are also very marketable, since many
outside the confines cf academia, sud, as news commentators, lobbyists
and political professionals, are interested in predicting election
outcomes, with economic conditions cften emerging as a very sclid basis
for a reliable political fcrecast.
General findings are consistent: Bad economic times speli trouble
for the partv in office. It did net take a generation cf research te confirm
what is clearly very intuitive, but it has taken a great deal of work te
identify the precise manner economics works. The conceptualizaticn of
economics and voting lias taken research in a variety cf directions,
leading to a varied set cf conclusions. The following review shows how
far economic voting lias gone to explain the link between economic
conditions and voter behavicur. But as will be seen, there is clearly far o
53 more distance yet to cover, and conventional economic voting lias, for
the most part, conducted research with a narrow focus.
Economic voting research confirms what is commonlv referred to
as the sociotropic thesis: Voters react to national economic conditions
ratlier than personal, or “pocketbook,” finances. Personal financial
considerations are nowhere nearly as important as national economic
conditions, findings that challenge the otherwise clean Downsian model
that sees voters adjust their support for the incumbent according to
personal economic circumstances (Downs, 1957). Egocentric effects do
exist, but tend to 5e weaker than sociotropic effects (Lewis-Beck, 1988).
Several reasons are given to explain why national conditions are more
influential than personal economic factors. First, it is flot aiways easy to
attribute blame or credit for changes in personal finances, but it is
possible to link general economic trends to an administration’s policies.
In some societies, such as the United States (the setting of most
economic voting research), an individualistic political culture reduces
the political relevance of individual circumstances decisions (Feidman,
1982, 1985; Sniderman and Brody, 1977). However, pocketbook effects
emerge more prominent when voters can clearly attribute personal
circumstances to government, a phenomenon particularly salient in
societies with more interventionist governments (i.e., welfare states),
where personal economic conditions are verv much tied to state policies
(Nannestad and Paldam 1994, 1995, 1997). C
54 Whereas the soci Lropic vs. egocentric debate pertains to the spatial context of economic voting effects, the retrospective vs. prospective debate focuses attention on the time dimension. Fiorina
(197$) was among the first to show that voters rely heavily on past economic conditions, given that past economic conditions are more readilv known. and therefore, blame and credit for past economic conditions are more easily established (Kinder and Kiewiet, 1979, 1981).
A formidable opposition to this view is contained in the forward-Iooking
“bankers” hypothesis (Erikson, et al., 2000; MacKuen, et al., 1992), which relates trends in consumer confidence in the United States to support levels for the President. findings substantiate a view of the voter as capable of assessing future economic expectations and linliing those expectations with the current governing administration. Other studies find room for both retrospective and prospective effects. Nadeau and Lewis-Beck (2001), for instance, note that retrospective effects are present if voters are evaluating an incumbent, such as the US. president seeking a second term, while prospective effects emerge more salient if there is no such candidate.
In either case, whether economic conditions are conceptualized as
retrospective or prospective, the time horizon is very short: about 12
months. Furthermore, some see this time horizon too long, where voter behaviour appear to correlate more strongly to economic changes a few months prior (Nannestad and Paldam, 1994). There is a good reason why it is not a good idea to extend the time horizon too far out. Much economic voting research is based on survey data, and respondents are
flot presumed to recali accurately economic conditions that go back
longer than a vear or two. Furthermore, it is not deemed reasonable to
expect respondents to project too far into the future. As a resuit,
economic voting research is based, for the most part, on the short-term.
Although survey-ievel researcli restricts the time horizon, the time
horizon of aggregate level work is flot much longer. One of the first
economic voting studies employs aggregate data (Kramer 1971), where
declines in aggregate-level income are related to reduced support for
incumbents, conceptualized as U.S. House of Representatives
candidates from the same incumbent piesidential party. The findings
reveal a pattern that is consistent with the retrospective hypothesis,2
where declines in economic conditions erode support from the
incumbent partv. But economic change is measured from one election
year to the year prior, a short-term horizon.
further methodological precision reveals other unique
characteristics of various indicators. Take unemployment, for instance.
High or Ïow unemployment rates may be politically important in
absolute terms, but what seems at least equally as important, if not
more so, are changes in the unemployment rate (Nadeau and Biais,
1993, 1995). Furthermore, voters seem to react to job growth levels if
2 Kramer (1983) points out that aggregate level data does net con6rm or negate an egocentnc hnk. Voting behaviour related to aggregate economic data cannot explain whetlier voters respond to changes to their own personal incomes, or to (j changes in national-level incomes.
56 their expectations are higher than actuallv reported by government statistics. Perceptions about unemployment are driven by personal experience, regional conditions, as well as partisan cues (Nadeau et al.,
2000). Others find interactive effects between economic and political factors. for instance, the clientele and “salient goal” hypotheses connect the relative salience of individual economic indicators to a party’s policy agenda. Here, unemployment and inflation mean different things to voters as they evaluate different parties. Some parties have a degree of
“ownership” over a particular problem (see, for example, Carisen, 2000;
Hibbs, 1987; Rattinger, 1991; Swank, 1993). A more recent stream of research focuses attention on regional conditions (see, for example,
Cutier, 2002; Godbout and Bélanger, 2002; Mondak et al., 1996), where conditions at the provincial, regional or even neighbourhood level have been measured to mediate sociotropic evaluations.
Overali, the different approaches to economic voting research yield interesting resuits. But two consistent features throughout the literature is the short-term horizon and the focus on support levels for the government (i.e., incumbent). What is needed is a look at whether voters respond to more structural changes, a more durable set of economic circumstances, the duration of which takes a much longer period to complete. If longterm economic decline chips away at support for both the incumbent and the mainstream opposition in favour of non mainstream parties, then a short-term horizon might not detect such a shift. Generally, support for non-mainstream parties is very small, and
57 changes over the shortterm are verv slight. Hence, a longer-term view is needed to measure something that otherwise might go unnoticed.
It is not to say that the long-term is completely ignored. Brooks and Bradv (1999), for instance, show that a longer-term view reveals voting behaviour as reflective of poÏicy evaluations. Here, voters evaluate partv stands on issues sucli as welfare programs, the size of government, the legitimacy of institutions, and other evaluations that require more than just 12 months to develop. There is nothing inherentlv economic about this category, given that policies address numerous non-economic issues. But in many instances, policies affect voters economicaliy, or are evaluated in light of economic conditions. As an example, Brooks and Brady mention the New Deal in the United
States as an income redistribution initiative that divided the
Republicans and Democrats. It is rational, then, for poorer voters to prefer greate; redistribution, while the wealthier prefer otherwise.
There is some evidence to suggest that a generation of painfuÏ economic restructuring and long-term economic decline can lead to voter resentment. Teixeira and Rogers (2000) point to 1973 as a dividing line in economic history, with the pre-1973 period marked by general affluence, while the post-1973 period was marked by a growing income gap between the upper and lower classes, a gap that has led to increased voter resentment. Sucli a perspective specifies sociotropic/retrospective evaluations as relevant before 1973, when ail hard-working Americans could expect to become middle-ciass. But the
58 stakes began to change after 1973, with a widening wage gap making long-term personal economic interests more salient among that segment of the voting population that was most adversely affected.
In sum, if voters see that no mainstream partv appears able — or
willing — to reverse long-term economic declines, then it should come as no surprise to see voters blame the “system.” But such a phenomenon cannot easily be captured by conventional methods that relate short term economic changes to opinions about the governing party. A one year improvement in economic conditions may certainly yield some benefit to the incumbent partv, but a lengthy period of stagnation miglit shift the focus of political discourse to the margins of non—mainstream political movements.
Conceptualizing Voting: Mainstream vs. Non-Mainstream
Economic voting studies that focus onlv on the level of support for incumbent parties cannot adequately measure levels of discontent.
Votes for the opposition Liberals during a PC reign is not the same as voting for the Ïeft-of-center New Democratic Party (NDP), or for Reform,
Bloc, Libertarian, and so forth. There is a different, and sometimes more aggressive, message communicated when voters support smaller opposition parties. There are different types of such opposition parties, each emerging under a unique context (see Gunther and Diamond
59 (2003) for a more complete and updated discussion), but it is sometimes not easy to categorize a particular political party into any ne type.
A common terni associated with a non-traditional opposition party is “third” party, a concept applied in Canada by Maurice Pinard.
According to Pinard (1973) a third party is simply a “non-traditional party which has not ye been in power,” (455). Frustrated voters who do flot see a satisfactory option among the traditional offerings would look for alternatives other than the Liberal and PC Parties. Both traditional parties emerged from within Canada’s parliament, making them essentially insiders, even when one party is in opposition. Whereas other parties have extra-parliamentary origins whose challenge to the two traditional parties can only be noticed by contrasting positions, which normally attack at the same institutions that have formed along with the traditional parties (Landes, 2002; Taggart, 1998).
Despite the simplicity of such a definition, the concept of a “third party” may be too specific for a study nf long-term economic voting. first, few third parties run candidates in all constituencies and in all provinces. The Bloc Québécois oniy fields candidates in Quebec. Othur parties, such as Reform, initially fielded candidates in the Western provinces only, and later expanded to other provinces except Quebec in
1993. Even then, not all constituencies in the “Rest of Canada” had a
Reform candidate. Second, even when a third party flelds candidates in ah provinces, its prominence and campaign strategy may vary from province to province, for instance, the NDP’s profile in Quebec has
60 aiways been negligible. Third, some of the most successful smaller
“third” parties in Canadian history have faded into oblivion, making a long-term focus difficuit. The Social Credit, Reform, Bc and NDP ail have captured the attention of those who wish to study voter discontent in Canada. But third parties (except, perhaps, the NDP) eventually fade away. Social Credit no longer appears on the Quebec electoral radar;
Reform has morphed into the Alliance which later merged with the PC party; the NDP has seen better davs. The rise and fali of these parties within a relatively short to medium time frame does not easiiy permit a study of iong-term voting trends. Fourth, the NDP might not qualify as a third party simpiy because it is not obvious whether this party is “non traditional.” The NDP may take a leftist stance on many issues, and may have had radical roots, but they are by no means a novelty in pohtics, and may simply be seen as a traditional opposition party (Gidengil, et al., 2001). They are nonetheless a party with extra-pariiamentary origins, and remain, at least federally, outsiders.
Even if it is feasibie to study a third party over a reasonably long period of time, the interpretation of votes for such parties wouid be open to question. One could argue that the Reform and the Bloc qualify as
“anti-state” parties. An anti-state party pursues fundamentai, at times even constitutional, change to reform the entire political system, not simply just to replace the current executive (Sartori, 1976). Communist and fascist parties fali into this category, running candidates in democratic elections just to acquire power, only to do away with
61 democratic institutions or impÏement drastic reforms — once in power.
Although flot necessarily as extreme as Communist or fascist parties, the Reform/Alliance party and Bloc Québécois could qualify as anti state. One party was founded to challenge standard Canadian institutions, such as the unelected Senate,3 and the other party was founded to pursue a different sort of fundamental change, Quebec’s independence. On the other hand, one could argue that neither party intends to uproot democratic institutions or to completely eliminate the
Canadian state, but instead hopes to rearrange and rebalance the institutions to reflect regional concerns.
New parties such as the Reform/Alliance party could also qualify as an “anti-party” movement. The basis of such movements stems from negative perceptions of established parties that lead to a view of parties as no longer able to represent the electorate (Poguntke, 1996). Declining turnout, dealignment and the rise of support for the Reform/Alliance party coincide with an increased sense of discontent towards Canada’s two main political parties (see, for example, Bélanger, 2004; Clarke and
Kornberg, 1993, 1996; Clarke, et al., 2000; Gidengil, et al., 2001).4 But
The Reform party’s populist views were also prominent. They demanded more “free votes,” the legislative recall, more referenda. Some of their more controversial views include opposition to official bilingualism and a general distaste for a centralized federation. 1 It should be noted that Poguntke (1996) distinguishes between specific and generalized anti-partyism. The first explains voter discontent with overali performance of governing or other key parties, whule the later points to voter discontent with the whole concept of a political party. Bélanger (2004) and Gidengil et al. (2001) show that sometimes a party can tap into both sentiments, as in the case of the Reform Partv of Canada, which appealed to
62 again, the Reform/Alliane partv does not Held candidates simply to oust the outsiders, but lias evolved to beliave less and less as a western protest partv and increasingly as a party uitli an eye on government, especially following the merger with the PC party.
While it is not aiways obvious whether a larger third partv reflects either anti-state or anti-party elements, the smaller parties are even less clear. What about the Green party, Christian Heritage, or “fringe” parties that at times appear to parody the political system (e.g., Rhinoceros)? To some extent, such smaller parties reflect both anti-state and anti-party sentiments. The more religious movements miglit derive support from those who regard mainstream political parties as too decadent, especially in light of hot-button issues such as abortion and homosexuaÏ riglits. More policy-oriented movements might reflect an anti-system views, with the Greens deriving some support from ecocentrics.
Given ail these particular constraints with the concept of a third partv, a more general concept is needed. The term “non-mainstream” party seems to conjure the same type of characteristics normaliy ascribed to third parties: an alternative choice. But “non-mainstream” is a more general concept, and non-mainstream voting can be defined as
simply the aggregate of votes cast for such alternatives. Table 1 . 1 lists ail non-mainstream parties that ran candidates during the period of this study. Size does not matter, because non-mainstream parties couid
those who were upset witli both the Liherals and Conservatives, as weli as those who otlierwise feit disengaged from partv politics altogether.
63 range in size from larger movements that have formed the officiai opposition, such as the Reform/Alliance partv and Bloc Québécois, to smaller parties and iess prominent candidates that hardly ever get mentioned, such as independents, the Marxist-Leninists, the Green party, Libertarians, Natural Law, and so forth. Ail non-mainstream parties are unique in what thev stand for and how they attract voters, but one could argue that the common feature among ail non mainstream parties is that a vote for such parties is ciearly flot a vote for politics as usual. A vote for a non-mainstream party is a rejection of mainstream politics.
When viewed as a whole, it is clear Canadians have voted in sufflcientlv high numbers against the two mainstream parties, the
Liberai and Progressive Conservative Parties. Prior to the watershed
1993 eiection, the proportion of votes cast for non-mainstream parties
hovered within a range 20 to 25 percent (see Figure 1.1), due mainÏy —
but not exclusively — to support for the NDP. Since 1993, support for the
non-mainstream parties surged past 40 percent, due mostly — again, not
exclusively — to the Reform/Aliiance party and the BQ. In other words, whiie support for particular non-mainstream parties is volatile over time, support overali for ail non-mainstream parties tends to be more steadv, and support Ïeveis tend to reflect a generalized expression for an alternative voice, sometimes based on rage, sometimes on ideoÏogy, but generaliy based on a rejection of mainstream politics as represented bv the typicai parties that tend to occupy government.
64 o
Table 1. 1: Non-mainstream parties that ran candidates in elections from 1979 to 2000
Abolitionist Party of Canada Bloc Québécois Canada Party Canadian Action Party Canadian Alliance Christian Heritage Party of Canada Communist Party of Canada Confederation of Regions Western Party Co-operative Commonwealth Federation Green Party of Canada Independent Libertarian Party of Canada Marijuana Party Marxist-Leninist Party National Party of Canada Natural Law Party of Canada New Democratic Party Parti Nationaliste du Québec Parti Rhinocéros Party for the Commonwealth of Canada Reform Party of Canada Social Credit Union Populaire
Source: http://www.parl.gc.ca
o
65 o figure 1.1: Vote shares of non-mainstream and mcumbent parties, 1979-2000 5Q%
10%
Q% -
1979 1980 1984 1988 1993 1997 2000 Non-mainstream Incumbent
Source: http://www.parl.gc.ca
Data and Methods
In order to link economic conditions to support for non
mainstream parties, provincial economic and provincially pooled federal
voting data were compiled. Canada’s 10 provinces offer neat packages of
both economic data and voting resuits, permitting the construction of 10
data points for every election being studied. Election resuits from
previous elections are conveniently and freely available from Canada’s
Library of Parliament web page. Ballots cast for every candidate,
whether affiliated to a party or not, are contained in these databases.
Aggregated provincial-level election resuits are used to construct two
separate variables. One variable tracks the percentage of votes cast for
66 non-mainstream parties, comprised of votes for candidates not affihiated
with either the two mainstream parties. Another variable tracks the
percentage of votes cast for the incumbent party. The Liberals were the
incumbent for the 1979, 1984, 1997 and 2000 elections; the PC was the
incumbent for the 1980, 1988 and 1993 elections. for the seven
elections covered, more than 87 million valid votes have been cast, the
vast proportion of which for the Liberal and Progressive Conservative
party, Canada’s two mainstream parties.
Combined to the voting data, provincial-level economic data from
1969 to 2000 were gathered from Statistics Canada’s Labour force
Suroey databases (CANSIM Table 282-0002), producing a pooled
aggregate dataset of 69 data points.5 Two key economic indicators were
extracted: unemployment rate, and participation rate. The
unemployment rate requires little justification, given that it is a valid
and widely used measure of economic conditions. The other labour
market indicator, the participation rate, appears neyer to have been
used in voting research. Economic research, especially labour
economics, does pay a great deal of attention to participation-rates, a
measure that reveals economic conditions not accurately captured by
the unemployment rate.
First, labour-force participation is affected by longer-term
conditions, such as demographics. Immigration, the baby-boom
The total number of data points would have been 70, but unemployment data , for Prince Edward Island xvas missing for some years during the early 1970s, reducing the sample size to 69 for two of the three models tested.
67 generation and the ent of women into the workforce ail affect the supply side of the workforce. Data from Statistics Canada reflect some of these demographic changes. For instance, the maie-to-femaie ratio of the labour force declined from .30 in 1976 to .23 in 2000.6 Second, labour-force participation may reflect overail economic health. Economic expansion, such as the period following the Second World War up to the
1960s. drew more peopie into the workforce, while trends from the
1960s to the 19$Os show the reverse. Economic shocks (e.g., OPEC), stagflation, downsizing and the outsourcing of manufacturing and industrial work to the developing world ah contribute to overail economic malaise that affects the “demand” side for labour. Industry automation can permanentlv eliminate the need for certain workers, rendering them economicaliy obsolete, with the consequence of seeing
larger segments of the population — especialiy those who lack a post
secondary education — disqualified from many of the growth areas of the economy, or qualffied only for work that does not pay as much as the blue-cohiar union wages of a previous generation. Such conditions appear to have negatively affected labour-force participation (Hoizer,
1990), which may explain tlie post-1970s siowing trend of the participation rate in Figure 1.2.
6 CANSIM Table 282-0002.
68 Figure 1.2: Labour market trends in Canada, 1969-2000 70% 15% Q) 6Do 1UYo
C D 0 50%- 1970 1980 1990 2000
Participation Unemployment
Source: Statistics Canada
If it persists, extended periods of economic decline discourage
workers who simply give up looking for work, while potential workers —
sucli as students who have recently graduated — do not even bother to
look (Sapsford, 1981). This might lead to unemployment, but it might
also lead to a different type of employment, such as part-time work.
According to Statistics Canada, the ratio of the part-time to fuil-time
workers grew from .14 in 1976 to .22 in 2000.
Ail of these factors contribute to overail earning potential, which
is another key factor that drives participation rates (Parsons, 1980).
Deciining income prospects discourage potential workers from seeking
employment. This degree of economic malaise is not aiways adequately ci Ibid.
69 captured by simply relving on unemployment figures. Participation rates are then a more comprehensive indicator.
Another good reason to use the participation rate is its ability to tap into long-term economic trends. Short-term dechne, incÏuding the occasional recession, may have some impact on participation rates, but the impact seems more pronounced on unemployment, which tends to bounce back more quickly. As shown in Figure 1.2, participation rates, on the other hand, increase or decrease independently of the business cycle. This suggests that the decision to join or withdraw from the labour force depends on long-term factors rather than short-term
economic swings (Ostry and Zaidi, 1979). In other words, the labour- force participation rate functions as a general indicator of overali
economic strength.
— For each main indicator — unemployment and participation two
different types of variables are created, one for short-term changes and
one that covers the long-term. Short-term changes are simply an
arithmetic subtraction of election-year rates less the rates of one year
prior. Long-term changes are calculated in a similar fashion, current
year rates minus the rates 10 years prior.
Canada’s political landscape cannot be fully understood without
taking into account regional considerations (Gidengil, et al., 1999),
hence the decision to include regional dummies. furthermore, since the
dataset comprises of provincially pooled data, and given that regression
analysis is being used to test the hypotheses, regional dummies are
70 necessa to account for the possibility that residuals miglit not be
independent across the regions. Three such regional dummies are
constructed; one for the Western provinces; another for Quebec; and
finallv, one for Atiantic provinces.8 Ontario was the excluded region and
thereby functions as the reference group.
The four economic variables (short- and long-term versions of
unemployment and participation) and the three regional dummy
variables form the basis of Ordinary Least Squares (OLS) regressions
moUds to test two different hypotheses. The first hypothesis tests the
responsibility hypothesis. Confirmation of this hypothesis would be
substantiated if short-term economic conditions explain support levels
for the incumbent party. 0f the two economic variables, the short-term
change in the unemployment rate would be expected to yield the most
significant resuits in explaining incumbency vote.
H1.1: Support for the incumbent is inuersely related to short-term
changes in unemployment.
The participation rate is not expected to score significant results, and
neither economic variable’s long-term versions is expected to yield
signfficant resuits.
H Western provinces include: British Columbia; Alberta; Saskatchewan; and Manitoba. Atlantic provinces include: Newfoundland and Labrador; Prince O Edward Island; Nova Scotia; and New Brunswick.
71 The second hypothesis tests the link between long-term economic
decline and support for non-mainstream parties.
H1.2: Support for non-mairistream parties is inuersely retated to
long-tenn changes in the labour participation rate.
The short-term version of participation rate and both versions of the
unemployment variable are expected to yield insignificant findings.
Resuits
The economic variables and regional dummies produce regression
models summarized in Table 1.2. Overail, the models appear stable,
showing no major violation of any assumption to a linear regression
model.’° The first model confirms the responsibility hypothesis, but with
mixed resuits. Overail, support for an incumbent could be explained by
short-term changes to unemployment, as hypothesized. However, short
term changes to the participation rate also yield a marginally significant
result (p<.1O), suggesting that it, too, should be included in the model.
See Appendix 1 for a compiete detailed description of ail variables. 10 Collinearity diagnostics suggest no problem witli respect to the reiationship among the different independent variables. Residual analysis also does flot suggest any other serious problems, except for Modeis 2a and 2b, where heteroscedasticity was detected. This was treated using a base-10 iogarithmic C transformation of the dependent variable.
72 o
Table 1.2: Comparing long- and short-term effects
1 2a 2h Dependent variable, Incumbent Non- marnstream partles* Vote share for: Std. Std. Std. Coef. Error Coef. Error Coef. Error Economic conditions, Long-term: ParUcipauon rate .233 ,455 -2.783 ,788 p<.001 -2.931 .637 p<.00l Unemployrnent rate -.367 .576 -.148 ,998 Economic conditions, Short-terni: Parucipation rate 4.679 2,651 p<.lO 1.215 4,592 1.748 4,534 Unemployment rate -3.831 1,576 p<.05 -1.343 2,730 -1.707 2,606 Regional dummies: West -.088 ,040 .250 ,070 p<.001 .254 ,067 p<.001 Quebec .0194 .064 .0862 ,111 .0881 ,109 Atiantic ,001 .039 -.178 .068 p<.OS -.185 ,067 p<.001 Intercept .324 ,038 p<.001 -.506 .066 p<.00l -.508 ,065 p<.00l R-sq .309 .626 .635 Adj. R-sq .229 .584 .601 SEE .105 .182 .181 N 69 69 70 tThe dependent variable (percentage vote for non-mainstream parties) was converted to a base-10 logarithm to correct for heteroscedasticity. o Figure 1.3a: Participation rate ami non mainstream vote, Western provinces 75%
r = -.716 (p<.001) s
V cV ‘ ‘ 25% V >
0% - - -2% 0% 2% 4% 6% 8% 10% 12% 10-year change in the participation rate
Figure 1.3b: Participation rate and non mainstream vote, Central Canada 75%
U,
D 4-
r = -.629 (p<.O5)
U, ci, 25% s s iEH >
0%- - — -2% 0°/b 2% 4% 6% 8% 10°,’ 12% 10-year change in the participation rate
Q
74 o Figure 1.3e: Participation rate and non mainstream vote, Atiantic provinces 7 5°/
1.)
( 5Q% o • r = -.401 (p<.01)
25%
• •4. • • > • • •• •• • 0% -2% 0% 2% 4% 6% 8% 10% 12% 10-year change in the participation rate
Models 2a and 2b confïrm the second hypothesis. Model 2a
clearly shows that the only significant economic variable that explains
support for non-mainstream voting is the long-term change in
participation rates. None of the short-term variables yields significant
scores, and neither does the long-term change in unemployment. (Model
2b exciudes long-term unemployment for reasons to be explained
below.) The second hypothesis is further confirmed visually through
regional-level scatterpÏots (see figures 1.3a to Y.3c). Ail three major
regional groupings show a consistent pattern: Weaker growtli in
participation is related to stronger support for non-mainstream parties.
In a sense, these resuits simplv suggest that long-term economic
decline does not boUe well for mainstream parties. This interpretation cD follows the conceptualization of labour-force participation as strictly
75 economic However, the participation rate could also be interpreted as measurement of overail integration into mainstream societv. Drops in labour-force participation reflect worsening long-term socioeconomic conditions. In sucli a context, a decline in the participation rate captures a very real sense of economic hardship, and might also reflect
“social exclusion” (Brady, 2003: 723). The decision to stop looking for work is a reaction to bad economic conditions, alienation, and an overali sense cf exclusion, sentiments normally exploited by non-mainstream parties.
However, as already pointed out in reference to Pinards work, the socially excluded tend to be non-voters. Hence, there is nothing to these findings to suggest that the inverse relationship between labour force participation and support for non-mainstream parties derives from voters who have withdrawn from the workforce. Instead, it may be those workers left behind to do ail the work that may feel resentment. In addition, they may not necessarily express their resentment by pointing to economic conditions. As Teixeira and Rogers (2000) indicate, rising discontent among working-class American voters stems from a sense of unfairness. The large segment of voters who work, pav taxes and struggÏe to make ends meet see a political establishment more intent on addressing the needs of minorities, welfare recipients and other citizens who otherwise do not work, and who probably do not vote, eitlier. But working-class citizens do vote in greater proportions, and the political relevance of their sense of unfairness grows amid worsening economic
76 conditions over a long period of time. It would then corne as no surprise to see such voters gravitate increasinglv towards political parties that advocate a social and political agenda wrapped in the language of worker-rights, taxpayer-rights and government waste. The political behaviour itself may be non-economic, but a drive for such a movement may very well stem from economic strains.
The other main economic indicator, unemployment, did not produce nearly as consistent a picture. Long-term increases in the unemployrnent rate are not accompanied by increased support for non mainstream parties, mostly because of the cyclical pattern of unemployrnent. It is clearly more volatility than the steadier pattern of participation rates (see figure 1.2). Therefore, a 1O-year change in unemployrnent rates might not accurately capture a trend. It is precisely for this reason that two different long-term models were constructed, one that included long-term changes to unemployment, and one without unemployment. Resuits of both versions do not differ by mucli, but it might make more theoretical sense to exclude unemployment, at least the way it is used in these models.
Among tlie regional variables, the West regional dummy is consistently significant for both models. The West dummy is negative in
Model 1, which suggests lower levels of support for the government within western provinces, the flip-side to the positive coefficient score in
Models 2a and 25. This is not overly surprising, given that non mainstream parties have often llared up in the West. The Progressives.
77 Social Credit, the New Democrats. and, rrare recentlv, the
Reform/Alliance parties have traditionally drawn a lot cf support from the West. “Western alienation” is a sentiment that clearlv has contributed to the success cf non-mainstream parties in that region. But would not economic dechne further enflame a region with a political context alreadv predisposed to view conventional Canadian politics as biased in favour of “the center?” In other words, economic decline in other parts of Canada might be interpreted as either bad economic decisions by state leaders or simply an ebb in the economic cycle, whuïe in the West it might be seen as a rip off. Lipset (1959/ 1963) and Pinard
(1971) both mention this possibility, and the data appear to validate them both. This may explain why the scatterpiot for the Western provinces (Figure 1.3a) shows a stronger inverse relationship between labour-force participation and support for non—mainstream parties than the scatterplots for the other two regions: The West may be the region that is most sensitive to economic changes, and consequently, more prone to transfer economic discontent to support for a non-mainstream party.
As for the other two regional dummies, one or botli cf the remaining regions fail te attain significance. Quebec neyer yields a significant resuit, probably due to the small sample size that accounts for this regional dummy. The Atlantic regional dumm is significant only for the model that explains vote for non-mainstream parties, and curiously, living in what is clearly Canada’s poorest region remoues vote
78 share from non-mainstream parties. One explanation for this nding is that the context ir the Atiantic provinces might be totally opposite to that of the West, whereby voters in the Maritimes are more likely to sustain their support for mainstream parties. Again, this validates
Pinard and Lipset, where povertv (and the Atiantic provinces are persistentlv the poorest in Canada) does not automatically cause people to revoit. Nonetheïess, even in the poorest region of Canada we see a significant inverse relationship between labour-force participation and support non-mainstream parties.
Conclusion
Resuits presented liere lead to several conclusions. First, the responsibilitv hypothesis is easily replicated, where support levels for the incumbent party could reasonably be tied to short-term economic changes. In particular, a drop (increase) in the unemployment rate over a period of one year boisters (hurts) voter support for the incumbent party. But it is one thing for voters to “kick out the rascals” and place in government another mainstream party, and quite another when voters increasingly prefer parties with more radical agendas. This leads to the second main conclusion: long-term economic decline appears linked to the support for non-mainstream parties. This finding lends support to
Lipset’s (1959/1963) theory that relates governmental legitimacy to
79 generalized affluence. If an increase in voter support for non mainstream parties is understood to mean a vote against the status quo as well as a vote in favour of a set of ideas that pursue fundamental changes, then the findings reported here support Lipset’s theory.11
Such resuits could not have surfaced bv relying on conventional economic voting techniques that focus on the short-term. The short term cannot easily detect any growing disconnect between voters and mainstream politics, an evolutionary process that expectedly takes a considerable amount of time. If voter shifts are presumed to move at glacial speeds, then short-term economic indicators can onlv tap into changes in support for the incumbent party. This is another important conclusion to draw from this article’s analytical approach: Long-term changes to economic conditions matter at least just as much as short term.
But this is as far as aggregate data can substantiate. The resuits presented here are based on a compilation of provincially pooled numbers, which cannot theoretically be extrapolated to individual behaviour. Theoretical validation requires analysis at the next level, namely analysis at the individual-level data. Election surveys and other micro-level data miglit form a basis upon which to construct such further studies, assuming appropriate data exist.
‘‘ Given potential ambiguity as to whether the NDP is a true non-mainstream partv, it should be noted that the resuits are generallv replicated if the NDP is excluded from analysis. However, this causes the standard error of the estimate to grow considerably. Similarlv, results remain unchanged if the marginal parties are excluded. This is discussed further in the Conclusion on page 162.
80 Purthermore, the conclusion that long-term economic decline increases support for non-mainstream parties may be a phenomenon unique to plurality systems. The same conceptualization of voting behaviour may not yield the same resuits in proportional representation systems. In PR systems, as clearly stated by Duverger (1954), voters who wish to support smaller parties do not necessarily feel their votes would be “wasted.” Consequently, voters cast ballots for parties that appeal to specffic political agendas, an effect that in turn encourages the proliferation of more parties. Hence, PR systems are more likely than plurality systems to encourage voter support for radical, separatist, regional, single-issue, extremist or other non-mainstream political parties, regardless of economic contexts. But in a plurality system, it takes a great deal of strife (such as long-term economic decline) to encourage voters to move away from the traditional ehoices and to take a chance on a non-mainstream party. Also, as Pinard (1971) notes, voters suffering strain shift their support to a third party when the main opposition party is weak in their constituency. In Canada’s case, the weakness of the Progressive Conservatives in rural Quebec led otherwise
Liberal supporters to vote for the rising Social Credit. Third parties in sucli a context can succeed locally or regionally but not nationally. But voters in a PR system do not require the context of one-party dominance to feel inclined towards supporting a non-mainstream party, since there a vote cast for a smaller party would not necessarily be wasted. Such an institutional context limits Pinard’s approacli, and also limits the extent
81 to which the resuits from this article could be generalized to different electoral systems.
In addition, PR systems further complicate matters through their propensity to form coalition governments. Such partnerships may make it difficult to separate mainstream from the non-mainstream parties.
Take Denmark and Austria as examples. Denmark’s 2001 election produced a coalition of the Liberal and Conservative parties, two parties that may be considered mainstream. But Denmark’s coalition also relies
on an opposition party, the far-right Danish People’s party, to govern with a majoritv. Austria, too, experimented with a coalition involving a far-right party, the freedom party. Should sucli non-traditional and otherwise unconventional parties be considered mainstream once thev are admitted into a governing coalition, even if such parties advance an ambitious and arguably non-mainstream agenda?
Multï-partvism and institutional contexts that favour coalition governments make the mainstream/non-mainstream conceptualization of voting behaviour difficuit to apply. More precision mav be required to relate long-term economic decline to increased support for whatever qualifies as “non-mainstream” in such contexts. Lately, that behaviour appears tied to the recent success of nationalist and far-riglit parties that challenge the status quo. It would be valuable to assess a link between growing support for such parties in other industrialized economies and the considerable economic restructuring that has taken place over the past generation. Neo-conservative challenges to the
$2 welfare state, OPEC o1 shocks and the shift of manufacrnring jobs to
developing countries may have ail ied to more challenging economir
conditions for voters. If the Canadian case is not unique, then 1ongterm
economic change and economic decline in other countries mav shed
light into the growth of political movements that challenge mainstream
politics.
o
83 Q
CHAPTER 2
L0NG-TERM EcoNoMIc DECLINE AM0NG
OCCUPATIONAL GROUPS: POLITICAL EFFICAcY AND
SUPPORT FOR N0N-MAINsTREAM PARTIES
This article is under review at the Journal ofCanadian Studie.
84 Democratic Choices: Turnover or Rage?
In a democracy, no elected officiai assumes tenure; turnover is expected. Even good politicians often fail to get re-elected, and that is
perfectly appropriate, since democracies empower voters to pick — and
remove — their leaders. But there remain questions as to the criteria voters use to make such a choice. Among the many different theories
that seek to expiain voting behaviour, perhaps the most robust findings emerge from those that iink voting behaviour to economic conditions.
Out of plethora of articles and books written about economic voting, the basic conclusion is actually quite simplet Good economic conditions encourage voters to support the incumbent, whule economic hard times lead voters to turn away from the governing party. Standard economic voting research produces consistent findings that by now there shouk be no dispute about the centrality of econornics in a political system.
This is nothing new. As noted by Lipset (1959/1963), even the
Ancient Greeks recognized a link between economic prosperitv and support for political leaders. Voters and politicians are fully aware that bad economic times yield potentially negative consequences for the ruling incumbent. But normally, the incumbent’s loss cornes at the victorv of the main opposition party.
In Canada, the ruling incumbent and the main opposition have tended to be either the Liberal or Progressive Conservative Parties. Since
Confederation in 1867, no other party lias become government. But over
65 the last generation, voters have been disengaging from these mainstream parties. Actually, more and more Canadians have been disengaging from politics altogether. Fewer Canadians turn out to vote, and among those that do, an inc’easing number have been gravitating towards those parties that voice a more radical agenda.
Standard-economic voting research does not account for such a phenomenon. It fails to link economics to more fundamental aspects of governance, namely stability and legitimacy. Economics here is not simply seen as an evaluative component for voters to judge whether the government is doing a good job, but is understood as the basis upon which voters evaluate whether the overail system is seen as legitimate.
Therefore, while voters expectedly turn away from the governing party amid economic decline, where do they end up? Do tliey shift towards political movements that raise questions flot oniy about particular political leaders or particular policies, but also about the way political decisions are made? To use Eastonian terms (Easton, 1975), does economic decline lead voters to focus attention on more diffuse political objects, such as the general principals and institutions that underlie governance?
Some of the more successful political parties in recent Canadian politics are those non-mainstream parties that appear to stand for fundamental change and that appear to challenge the established principles of the Canadian model of governance. The Reform Party emerged in the Western pruvinces to turn Canada’s appointed Senate
86 into something that resembles verv much the American Senate: elected members with equal representation from each province and empowered with greater legislative authority. This, it is beiieved, would rebalance the federation, which is presently viewed as biased towards the vote-rich central provinces. The Bloc Québécois emerged in 1990 after the collapse of the Meech Lake constitutional accord and has since been asserting reform to the federal system in favour of Quebec independence. Even the not-so-recent New Democratic Party has a longstanding reputation of pushing for fundamental reforms, be if nationalization of corporations, socialization of health care, and electoral reform. And then there are the many smaller parties. Some appeal to more ideologically pure visions (e.g., the Marxist-Leninist Party), some to issue-specific platforms (e.g., the Marijuana Party), others to fringe movements (e.g., the Natural Law Party), and a few may even be considered parodies (e.g., Rhinoceros Party). But most, if not ail, of these non-mainstream parties appear to draw support from voters who feel a need to address a sense of unfairness and imbalance r garding the
Canadian configuration of power. When voters mobilize around such a theme, they are not looking simplv to change individual political leaders at the heim, but instead are pointing their criticism to the overali way that they are being governed.
Traditionaliy in Canada, such an agenda is raised by smalier parties. And the pluraiity eiectoral formula would tend to downpiay the significance of these parties, since the system favours larger parties. Yet,
87 Canadians have voted for these smaller parties for many decades, albeit flot in large numbers, until 1993. But over time, support for smaller parties lias varied. What explains this variation? Is variation in support just random, or is it linked to something more structural? These questions cannot easily be answered by standard economic voting research. The “responsibility hypotliesis” that forms tlie basis of e onomic voting models focuses on “specific” objects. But people experiencing economic decline may reject not only the party that happens to govern; tliey may also begin to question tlie whole political system. Voters suffering long-term economic decline, especiallv if they identify with a group that can be considered as consistentl disadvantaged, may re-examine their political loyalties framed as “us vs. them.” In otlier words, if voters identify witli a group (Nagler and
Niemann 1997; Nagler and Willette, 1999) and realize tliat the economic experience of the whole group lias not fared well over the long term, then members of this declining group may begin to question the system’s fairness and legitimacy. Do sucli voters simply shrug off their apparent disadvantaged conditions as simply part of the ups and downs of ordinary life in an advanced industrialized society? Or, do such voters interpret their peers’ suffering as unjust? Do tliey begin to question tlie integrity of the political system? Would sucli a group begin to manifest negative attitudes towards the system?
from this perspective, it is useful to link the Iong-term economic experience of occupational groups witli voting behaviour and with
88 attitudes about the political system, nameÏy external political efficacv.
External political efficacy measures the extent to which a citizen
perceives the political system as functional, or, more precisely, as
responsive. If the political svstem appears to be playing favourites, such
that some groups appear to win while other groups consistentlv lose,
then the loser may regard the system as non-responsive. If non
responsive means being an “out group,” then long-term economic
decline would most definiteÏy qualify for outsider status. Voters in such
a category would not only register lower levels of external political
efficacy, but they would also support parties that echo these sentiments
of discontent. But in order to understand such a phenomenon, a more
long-term and broad perspective is required to connect economic
changes, attitude changes, and fundamental voter shifts.
Narrow Focus of Standard Economic Voting Research
Economic voting’s researcli predominant focus is the short-term.
A common method of data gathering is to use surveys to gauge voters’
perceptions of the economy over the previous 12 months. ] This
approach is widely successful, able to generate accurate predictions of
voter support for an incumbent. This approach is also based on the
‘ Voters can also be seen as forward-looking, or prospective. See Erikson, et al., Q 2000, and MacKuen, et al., 1992.
89 voter myopia assumption, where voters’ memories are assumed not to
extend longer than a few quarters (Nannestad and Paidam, 1994).
Economic conditions that extend past one vear are seen as too distant to enter the voting caiculus.
Alternatively, analysis could rely on actual economic data instead
of survey responses. But even here, economic voting studies tend to rely on short-term data. Kramer (1971, 1983) was among the first to relate voting behaviour to real-world economic data, with resuits validating the responsibility hypothesis: Deciines in economic conditions in the United
States one year before an election erode support for House of
Representatives candidates of the president’s party.
Another trait of standard economic voting research is its tendencv to view voters as one homogenous group, as if ail voters share a similar economic experience. 2 Ail voters are assumed to know whether the country as a whole is growing or in decline. But economic changes are not experienced the same way throughout the economy. Sectors differ according to the extent of change (some sectors grow more than others) and the direction of change (some sectors grow, others suffer). The question is: Do voters identify witli their “sector?” There is some suggestion that voters do, indeed, evaluate the economy in light of reference groups. Within standard economic voting, some have attempted to disaggregate voters. Weatherford (1978) categorizes voters
2 Some studies take into account voter heterogeneitv with respect to sophistication and knowledge (see, for instance, Krause, 1997).
90 according to class. Others (see, for example, Carisen, 2000; Goodhart
and Bhansali, 1970; Hibbs, 1987; Rattinger, 1991; Swank, 1993) note
how some voters respond more to inflation, whule others to
unemployment. Voters are also disaggregated according to region
(Cutier, 2002; Godbout and Bélanger, 2002; Mondak et ai., 1996). And
others have argued for examining the iink between economics and
voting through a voter’s “economic reference group,” such as one’s
occupationai group (Nagler and Niemann, 1997; Nagler and Willette,
1999). If a voter’s economic reference group is doing well, then that voter
would tend to support the incumbent.
Another constraint with standard economic voting research is its
tendency to interpret voting behaviour as a zero-sum game between the
incumbent and its main challenger. This perspective may reflect the fact
that much economic voting research takes place in the United States,
where ‘oters have only two viable choices, the Repubiican and the
Democratic Party. But in Canada and most other countries, voters do
indeed have more choices. In fact, some parties are clearly associated
with the system while others are more peripheral. Should a vote against
the ruÏing Liberal Party be seen as the same as a vote for the Bloc
Québécois, or a vote for the Greens, or even, for simply not voting at ail?
Standard-economic voting, in generai, has little to say about ail these
different choices, and this oversight weakens its interpretative leverage.
Voters who turn their backs on mainstream politics punish the ruling .
91 incumbents, but they also jeoparJize the stability of the entire political
system.
Non-Mainstream Parties
The Liberals and Conservatives dominated Canadas parliament at the founding and for the remainder of the y9th century and are therefore seen as the two parties most connected with the overali political system. Anger against the overail political system is understood to implicate both of these parties. In other words, a vote for either mainstream party is seen as one unique behaviour that is reflective of ordinary, stable politics, while a vote for a party other than any of these two mainstream parties is seen as an expression for something compÏetely different. Therefore, if voters wish to express anger agaÏnst the entire political system, voters are expected to shift support away from the mainstream and towards non-.mainstream alternatives.3
Canadian history is dotted by the rise of non-mainstream movements that challenge the status quo. Parties such as the
The concept of “third party” miglit also capture the same type of voting behaviour. Pinard (1973) defines a third party as a “non-traditional partv which bas not yet been in power,” (455). Although “third party” seems generic enough, such a concept seems more apt to explain the rise of a party that is normally a regional phenomenon, and a party that, effectively, finishes in third place. If flot third, then maybe a close fourth. Thc Creditistes’s success in Quebec and the Reform Party’s emergence in Western Canada are two appropriate examples of “third parties.”
92 Progressives,4 Social Credit, and more recently, the Reform Party and
Bloc Québécois have garnered support from those who feel the entire political system is in need of fundamental reform. Perhaps one of the most successful non-mainstream parties is tlie Co-operative
Commonwealth Federation, which later became the NDP. This traditional opposition party (Gidengil, et al., 2001) lias extra parliamentary origins and remain, federally at least, political outsiders.5
But Canadians have voted for more than just the more prominent parties among the non-mainstream offerings. More marginal parties have garnered votes, such as the Libertarians, the Marxists-Leninists,
Christian Heritage, and so on. In any case, Canada’s plurality svstem and the traditional competition between the two mainstream parties have made a choice to vote against either mainstream option cognitively charged. A voter must reflect upon the possibility that a vote for a non mainstream party may be a waste. But flrst, such a voter would have to determine whether it is worthwhile even to vote at all. In either case, the basis of such decisions is connected to an overali evaluation of the functionality of the political system.
The Progressives eventually disappeared, with many of its members joining the Liberals (see Morton, 1950), but its leader went on to head the floundering Conservative party in 1942, on condition that the party rename itself Progressive Conservative. That name stuck until the floundering PC Party merged with the Alliance Party (formerlv Reform Party), with the new entity calling itself the Conservative Party. Resuits of the analysis in this article are unaffected if NDP voters are excluded. Also, results are generally replicated if analysis excludes marginal parties, that is, parties other than the most prominent “third parties” (i.e.. the NDP; Social Credit for elections during the 1970s; Bloc Québécois from 1993 onward: and Reform/Alliance from 1988 onward). Please see the Conclusion (page 102) for furtlier discussion.
93 External Efficacy: Measure of System Support
A vote for or against the status quo depends on whether or flot
one regards the status quo as appropriate. Voters drawn to non
mainstream parties, lie they traditional opposition parties sucli as the
NDP, populist movements sucli as Reform, or one of the many more
obscure parties are assumed to support parties that are “outside” the
system. These parties formed outside Parliament, have neyer won a
federal election or been part of a federal government. In fact, the only
officiai coalition government in Canadian history was that of the two
mamstream parties. The Unionist government of 1917-1920 grouped
together Conservative MPs and Liberal MPs who were in favour of
conscription during the First World War. The NDP has, at times,
supported Liberal minority governments, but was neyer officially a part
of the government. In any case, non-mainstream parties are political
outsiders, and their outsider status is used to empathize with voters’
feelings that the svstem lias shut them out, and often put forward
radical solutions to change things. For instance, the Bloc Québécois’s
solution is for Quebec to leave Canada; communists want to eliminate
private property.
In political behaviour, one of the more commonly used indicators
of such an attitude is external politicai efficacy. It is a measure that o
94 dates back to the 1950s, wlien if was part of a five-item scaie6 ol
“efficacy” (Campbell, et al., 1954). Later studies (Baich, 1974; Craig and
Maggiotto, 1982; Lane, 1959) demonstrate a need to separate two sub
dimensions: internai efficacy, which measures the extent to which voters
see themselves able to impact the system, and external efficacy, which
measures the extent to which voters regard the political system as
responsive. More recent studies add further sophistication to the
concept. Craig et ai. (1990), for instance, distinguish not only internai
efficacy from external efficacy, but they also identify other related
concepts, such as regime-based trust and incumbent-based trust.
Despite issues related to its multidimensionality, efficacy is seen
as a set of attitudes that form through sociahzation (Easton and Dennis,
1967; Iyengar, 1980) and is therefore durable and resistant to change
(Aish and Jàreskog, 1990). It is an underlying basis of a society’s
political culture, and it enables us to distinguish more participant from
less participant orientations (Almond and Verba, 1963), with levels of
efficacy directly reiated to democratic modes of participation, sucli as
vote turnout (Abramson and Aldrich, 1982).
Despite its durability, levels of efficacy can vary among different
subgroups of any society. Inter-group variations can be explained by
6 The original survey items are: I) “I don’t think public officiais care much wliat people like me think;” ii) The way people vote is the main thing that decides how things are run in this country;” iii) Voting is the only way that peopie like me can have any say about how the government runs things;” iv) People like me don’t have any say about what the government does;” and y) Sometimes politics and government seem so complicated that a person hke me can’t really Q understand what’s going on.”
95 socio-economic status (Wu. 2003), and even by general econornic
conditions. For instance, Canadians who live in more “central” economic
regions tend to measure higher levels of efficacv cornpared to those who
live in more “peripheral” regions (GidengiÏ, 1990). But efficacy cari also
change over time. It may flot happen quickly, but if a regime fails to
deliver positive resuits, and if people fel that their input is less valued,
then it should corne as no surprise to find more and more citizens
adjust their attitudes accordingly. In particular, citizens who
increasingly regard themselves as shut out of the political process are
expected to witlidraw support from the regime. As stated bv Madsen
(197$): “While one should assume no exact correspondence between
efficacy and support, it nonetheless is surely true that when a
supposedly democratic system is seen to have faiÏed in its promise of citizen efficacy, it is likely to lose the support of its disillusioned members,” (86$, emphasis his).
Disparities in social and economic conditions provide voters with dues as to the success or failure of the political system. One group of voters experiences negative resuits and blames the incumbent government. But another group of voters experiences negative results but feels discouraged at the possibility of influencing any real change. It is this second group of unhappy voters that deserves more scrutiny, because it is this group that is most likelv to disengage from mainstream politics, either by withdrawing completelv, or by moving towards political movements that are potentially destabilizing.
96 Data and Methods
Analysis of the link between economic conditions and voter support is broken down into several key steps. first, voters are categorized according to occupational groups. Statistics Canada’s 1980
Standard Occupational Classification (SOC) system catalogues more than 400 specific occupational tities that are sorted into 15 occupational groupings. The economic indicator used to measure economic change is based on the notion of ‘job growth,” a measure which has not yet caught the attention of standard economic voting research (but see Lewis-Beck and Tien, 2004). Labour force data contained in the 1971, 1981 and the
1991 Censuses are used to measure the growth of each of the 15 occupational categories over two separate 10-year periods: 1971 to
1981, and 1981 to 1991. Economic health is measured through the 10- year percentage change of each occupational group’s workforce’s proportion in the overall labour market. This conceptualization assumes that an occupational group employing fewer and fewer people is experiencing economic decline. Table 2.1 reports these changes for each ofthe 15 occupational groups.
97 Table 2.1: Job growtli for occupational groups, 1971-1991 10-year change Percent of labour force (percentage) 81 Occupational Group 1971 1981 1991
Managerial/Admin 4.905 7.247 12.517 47.747 72.720 Natural Sc./Math 3.085 3.588 4.12 16.305 14.827 Social Science 1.042 1.679 2.277 61.132 35.616 Teaching 4.602 4.355 4.509 -5.367 3.536 Medicine/Health 4.303 4.622 5.235 7.413 13.263 Arts 1.06 1.473 1.767 38.962 19.959 Clerical 18.098 19.503 18.519 7.763 -5.045 Sales 10.748 10.205 9.419 -5.052 -7.702 Service 12.78 12.741 13.087 -0.305 2.716 Agriculture 7.106 4.884 3.769 -31.269 -22.830 OtherPrimary 1.666 1.396 1.055 -16.206 -24.427 Processing 4.411 4.202 2.956 -4.738 -29.653 Machining 11.533 11.016 8.314 -4.483 -24.528 Construction 7.491 6.852 6.07 -8.530 -11.413 Transportation 7.171 6.236 6.387 -13.039 2.421
Source: Statistics Canada. Occupation. March 1993. Cat. No. 93-327:
Statistics Canada. 1981 Census of Canada. Labour force - Occupations Trends. Nov. 1983. Cat. No. 92-920.
As can 5e seen, some groups seen its proportion ofthe workforce,
grow such as administrators, health professionals, and the large
“service” sector. A handful of other occupations were in decline,
represented by the bold scores. For instance, transportation workers
accounted for a smaller proportion of the workforce in 1981 compared to
1971, with the 10-year change amounting to 13 percentage drop. But
this group recovered somewhat from 1981 to 1991, with its proportion of
the overali workforce showing a slight 2.4 percent growth. C
98 Some occupational groups have even suffered two decades of dedline, such as in agriculture, and mining (i.e., other primary). nterestingly, it is these occupational groups that Lipset (1959/63) points to as potentiallv politically unstable, given their relatively insecure economic conditions. In the data presented here, even occupations in manufacturing (i.e. processing) have suffered long-term decline, probably due to the growth of the well talked about “service economy.”
Since the focus of this paper is to Ïink economic hardship to support for the political system, the 1O-year change in each occupational group’s share of the job market is recoded such that negative changes (declines) are retained but positive changes (growth) are given a score of O. Ail scores were then converted to their absolute
values. Scores range from O (no dechne) to positive values, such that the higher the score the more worse off an occupational group. This arrangement was done because the hvpothesis being tested is that economic hardship fuels a sentiment of political inefficacy and leads to rejection of main stream parties, but there is no expectation that growing occupational groups are necessarily more supportive of mainstream politics than groups who are relatively stable. This may very well be the case, but the focus of interest here is the difference between those who are suffering some sort of economic decline and ail others.
Each 1O-year change is related to data from a particular
Canadian Election Study survev. Tvpicallv, the CES tracks a
99 respondents occupation, according to the Statistics Canada’s SOC
scheme. This facilitates the matching of 1O-year changes for anv one
occupational group to survey data from a particular election. Voting
behaviour of respondents in the 1979 election study are explained by
their occupational group’s 1971 to 1981 change in workforce size
relative to the entire labour market,7 while voting behaviour of the 1993
election study respondents is explained by their occupational group’s
changes from 1981 to 1991.8 The first dependent variable, vote choice, is
structured as a nominal variable with the following three categories: 1-
Vote for mainstream party (Liberals and PC); 2- Vote for non
mainstream parties; and 3- Abstention.
The 1979 CES did not specifically track occupation, but it did record each respondent’s “Blishen scores” (see Blishen, 1987), which was converted into 1980 SOC codes. 8 Data from the 1979 Canadian Election Study were made available by ICPSR. The data were originallv collected by Harold Clarke, Jane Jenson, Lawrence Leduc, and Jon Pammett. Neither the ICPSR nor the original collectors of the data bear any responsibility for the analyses or interpretation presented here. Data from the 1993 Canadian Election Studv were provided by the Institute for Social Research, York University. The survev was funded bv the Social Sciences and Humanities Research Council of Canada (SSHRC), grant numbers 411-92- 0019 and 42 1-92-0026, and was completed for the 1992/93 Canadian Election Team of Richard Johnston (University of British Columbia), André Blais (Université de Montréal), Henry Bradv (University of California at Berkelev), Elisabeth Gidengil (McGill University), and Neil Nevitte (Universitv of Calgarv). Neither the Institute for Social Researcli, the SSHRC, nor the Canadian Election Team are responsible for the analyses and interpretations presented here. It would have been ideal to include other election surveys, but this was prevented because of compatibility issues regarding occupational categories. The 1981 Census followed the 1980 Standard Occupational Classification, and documents convenientlv reported comparative data for both the 1981 and the 1971 census periods. The 1991 Census also reports occupational data according to the 1980 SOC, but subsequent censuses converted to the 1991 SOC, which is not compatible with earlier versions. Also, while efficacv items were asked in the 1979 CES, these questions were not repeated in the 1980 D CES. The 1980 CES is therefore excluded from the analysis.
100 The 1979 and the 1993 CES also measure respondent’s levels of externai efficacv. The two survey items used to measure external political efficacy are: i) I dont think that the government cares mucli what people hke me think,” and ii) “Peopie like me don’t have any say about what the government does.”9 Agreement with these statements is assigned a lower score than disagreement, so that high scores pertain to higher levels of efficacv. Scores from the two-items were added together, then the sum was divided by two, yielding an overail a range from 0 (10w efficacy) to 1 (high efficacy).
In ail, it is expected that economic decline experienced at the level of occupations erodes an individual’s sense of external political efficacv, which in turn lowers the propensity to vote as well as the propensity to vote mainstream. This relationship can be summarized in the following hypotheses:
H2.1: Voters in occupationai groups experiencing lower leveÏs ofjob
growth are less likely to vote and more Ïikety to support non
mainstream pczrties.
H2. 2: Voters in occupatzonal groups experiencing lower levels ofjob
growth regard the politicaÏ system as less responsive.
° In the 1979 CES, the items are listed as variable 1042 and 1044. In the 1993 CES, the items were found in the mailback survev wave as variable MBSD5 and MBSD8. The externai efficacv index for the pooled dataset yields a Cronbach’s alpliaof.5819.
101 Given that external political efficacy is expected to function as an intervening variable, it is expected that when ail three variables
(occupational group job growth, externai pobtical efficacy, and voting behaviour) are combined into the same model, the economic variable
Uob growth) is expected to iose its signfficance. This is summarized by the following hypothesis:
1-12.3: The reÏationship between occupational job growth cari uoting
behaviocir disappears when externaÏ political efficacy is inctuded
as u coritrol variable.
The model aiso includes other control variables, mostly socio demographic factors. These are: region, 10 unionization, 11 gender, 12 religion,13 ianguage,14 age.
The hypotheses were tested using a dataset that pools the 1979 and the 1993 CES. This was doue to overcome the problem of small sample sizes. Given that few respondents support non-mainstream parties to begin with, a cross-sectionai election survey would include
10 Three dummy variables were constructed: one for the four Atlantic provinces (Newfoundland, Nova Scotia, New Brunswick, and Prince Edward Island), one for Quebec, and a third for the West (Manitoba, Saskatchewan, Alberta and Britisli Columbia). Respondents who live in one of the three regions are given a score of 1 on the appropriate dummy variable, O otherwise. Ontario was selected as the reference group. A score of 1 is assigned to a respondent who belongs to a union. 12 Male=1, female=0. ‘ Catholics assigned a 1, all others, including those wlio are flot affihiated with any other religion, a 0. 14 Francophones are assigned a 1, all others a 0.
102 even fewer such voters’5. Add to that the fact that many of these voters
do not supply valid replies to ail of the items used in the multivariate
analysis, then the probability of failing to find significant relationships ta
Type II error) is high. For this reason, analysis of non-mainstream voting is facilitated by pooling together both CES surveys into a larger dataset.
The unprecedented success of non-mainstream parties in the
1993 election may impose a bias on tlie results. Therefore, a special dummy variable was included to account for the unusuai nature of the
1993 election. In particular, respondents from the 1993 CES are given a score of 1, while those from the 1979 election are given a score of 0.
Resuits
The first test uses a multinomial regression model, with results reported in Table 2.2.16 The positive coefficient for the variable “10-year change in occupational growth” shows that the more an occupational group is in decline the more likely a voter will opt not to vote, and the more likely a voter will choose a non-mainstream party.
15 J the 1979 CES, 422 respondents indicated having supported a non mainstream party; the number rises to 679 in 1993 16 See Appendix 2 for a complete detailed description of ail variables used in this article.
103 Q Q
Table 2.2: Multinomial logit estimates of vote choice, 1979 & 1993 Canadian elections
Non-mainstream Abstain from voting Coef. Std. Err. Coef. Std. Err. Occupational growth .012 .006 p<.05 .032 .007 p<.00l Regional dummies Atiantic Canada -.609 .178 p<.01 .223 .195 Quebec .444 .163 p<.01 .264 .232 Western Provinces .732 .113 p<.00l .521 .158 p<.Ol Religion -.144 .114 -.191 .153 Language .220 .155 .044 .213 Age -.012 .003 p<.00l -.029 .004 p<.001 Gender .104 .094 -.007 .128 Union .271 .095 p<.Ol -.083 .134 1993 1.067 .095 p<.00l .429 .128 p<.Ol Constant -1.425 .181 p<.001 -1.071 .235 p<.001
Pseudo-R2 .073 Log likelihood -24 12.49 N 2909 Note: The baseline category is vote for mainstream party. Figure 2. 1 provides a visual simulation of this pattern. Tlie
effect, though flot very large, is nonetheïess noticeable, and contrnuous.
Support for non-marnstream parties and the probability of abstention
steadily increases as economic circumstances worsen. The shifts mav
appear siiglit, but in an election, a few percentage points either wav can
make the difference between winning a majority government, wiflning a
minority government, and losing entirely. So any small shift of support
away from the mainstream parties to any of the non-mainstream
parties, especiallv if a non-mainstream party lias a prominent regional
presence, can vield to completelv different electoral outcome.
Figure 2.1: Simulation of vote probabilities and occupational-level economic change .75
.50
ï I Mainstream Non-mainstream Abstain from voting
Decline of 25 percent D Decline of 10 percent • No decline
17 Figure 2.1 was generated with Stata using Clarifv, a macro deveioped by Tomz, Wittenberg and King (2003), and available at http://gking.harvard.edu/ stats.shtml. (See also King, Tomz and Wittenberg, 2000). Its purpose is to demonstrate graphicaliv the distribution of statistical models using simulation techniques. The charts show the relationship between the economic variable and the dependent variable, while setting ail other variables (the control xiariables) to their mean values.
105 There are other significant effects, but the regional variables deserve some attention. Western Canadians appear here (as in other studies of Canadian voting) to be more prone to support non mainstream parties. This is not surprising since Western Canada has traditionally been welcome to populist politics. The Progressives, the
CCF, and Reform are three examples of parties that succeeded in the
West. Between 1972 and 2000, non-mainstream parties garnered an average of 40% of the vote cast among the Western provinces. ‘
The resuit for the Quebec regional dummy is similar. Living in
Quebec raises the probability of voting for a non-mainstream party compared to living in Ontario. Quebec has a history of voting for either the incumbent (usually the Liberals) or for a non-mainstream party, such as the Créditistes and the Bloc, so this finding does not mn counter to expectations.
Atiantic Canadians are opposite to the West, both geographically and politically. Living in the Maritimes appears to lower the probability of voting for non-mainstream parties compared to either of the mainstream options. Again, this finding is no surprise, since the Atiantic provinces are widely known to support mainstream parties. Between
1972 and 2000, non-mainstream parties garnered on average 16% of the vote in the Atiantic provinces. This might at first appear curious, since living in Canada’s poorest region apparently does not harm support for the mainstream. But Lipset (1959/1963) and Pinard (1971) poïnt out
8 Source: http://www.parl.gc.ca.
106 that it is impoverishment, and not poverty per se, that leads one to
become politically volatile. At worse, the poor might simply disengage
from politics. But when one faces continued economic decline, then
support for non-mainstream parties becomes more palatable. The data
here bear this out. When the simulation process applied to produce
Figure 2. 1 is repeated for the Atiantic provinces, only, the probability of
abstention grows from about .10 for no decline in occupational growth,
to .16 for a decline of 15 percent, to .23 for a decline of 30 percent. In
comparison, the probabilities of abstention for Quebec, only, range from
.09 (zero decline) to .18 (30 percent decline). Ontario appears the most
participatory, with its abstention probabilities ranging from .08 to .17.
While for the West, the range is from .10 to .21.
The next set of results examines the link between the
independent variable (occupational job growth) and external political
efflcacy as the mediating step. In general, the mean score on the efficacv
index for the pooled dataset is .4, which can be considered low, given
the range of 0 (10w) to 1(high). But an OLS regression model suggests
that this already low level can drop further stili among those
respondents who work in declining occupations (see Table 2.3). o
107 Table 2.3: OLS regression of political efficacy, 1979 & 1993 Canadian elections
Dependent variable: Exerna1 Political Efficacy
Coef. Std. Err. Occupational growth -.003 .001 p<.00l Regional dummies Atiantic Canada -.0 18 .017 Quebec .036 .019 p<.lO Western Provinces -.023 .013 p<.lO Religion -.007 .013 Language -.015 .018 Age -.001 .000 p<.O1 Gender .021 .011 p<.lO Union .005 .011 1993 -.071 .011 p<.001 Constant .527 .020 p<.001
Adj-R2 .050 N 1751 SEE .220
Voters in occupational groups that have shrunk relative to the overail labour market tend to score lower in political efficacy, .40 compared a score of .45 among voters whose occupation have grown.19
This may not seem hke a huge difference, but it is worth pointing out the impact of economic decline yields a larger impact on external efficacy than the Western regional dummy variable.20 In addition, job growth is
10 These scores were calculated by dividing up voters into two groups: 1) those whose occupations have shown negative growth (i.e., decline), and 2) ail others. An mdependent samples T-test shows that the 11-point difference between the two groups is statistically signiflcant (t=4.666, p<.001, df= 1903). ° The “Occupationai growth” variable obtained a beta of -.095, while the Western regional dummy scored -.050.
108 one of the few variables that reached statistical significance at p<.O5 or
better. The others being age and the 1993 dummy.21
When the model shown in Table 2.2 is re-run with external
political efficacy as a control variable, the resuits confirm expectations
(see Table 2.4). Political efficacy mediates between the exogenous factor
of economic decline and vote choice. When the dependent variable is the
probability of non-mainstream voting compared to mainstream,
occupational decline completely loses its significance. This suggests
quite strongly that the link between long-term economic conditions and
vote choice is not direct. Voters who work in occupationai groups that
experience long-term economic decline are more likely to acquire more
negative political attitudes, and it is these attitudes that guide such
voters towards non-mainstream politics, as per Hypothesis 2.3.
But when the dependent variable is probability of not voting
compared to voting mainstream, economic change at the occupational level remains significant. As was the case in Table 2.2, respondents are less and Iess likely to vote as their occupational group shrinks in size relative to the overail workforce. furthermore, lower levels of external efficacy relate with high probabilities of abstention. In ail, these mixed results provide partial confirmation of Hypothesis 2.3.
21 The actual p value associated with the Western dummy to the nearest thousandth is .062.
109 C p<.lO p<.Ol p<.001 p<.O5 p<.00l voting Err. from .399 .196 .234 .189 .196 .354 .320 .006 .233 .011 .306 .416 Std. elections Abstain .151 .574 .238 .346 .773 .026 -.350 -.050 -.034 -.342 -.036 Coef. -1.756 Canadian 1993 & 1979 p<.001 p<.001 p<.001 p.<.O5 p<.001 p<.lO p<.05 party. Err. choice, .274 .004 .131 .147 .150 .125 .007 .230 .217 .122 .205 .263 Std. vote mainstream of Non-mainstream for .201 .093 .012 .326 .081 1631 .799 .829 Coef. -.283 -.012 -.482 -.674 1.171 -1.188 -1303.56 vote estimates is logit category efftcacy growth Provinces Canada Multinomial baseline political dummies 2.4: The Quebec likelffiood Western Atiantic 1993 Occupational Regional Constant Log External Religion Gender Union Pseudo-R2 Language N Table Note: Age Conclusion
The general findings validate a need to look at economic factors
over the long—terni, and a need Lo interpret voting behaviour not simplv
as support for or against the incumbent. Instead, a person’s long-terni
economic experience within his or ber occupational group influences
attitudes about the overali political system, and those attitudes affect
the propensity to vote, and among those that do, those attitudes affect
the propensity to support non-mainstream parties. These flndings
confirm both Easton’s assertion (Easton, 1975; Easton and Dennis,
1967) that worsening conditions erode citizen support for more diffuse
political objects, such as the regime, and Lipset’s theory (Lipset,
1959/63) that economic decline leads citizens to question the legitimacy
of their political institutions.
But why would someone in such a situation blame the
government? Why not simply attribute an occupation’s demise to global
economic factors? One possible answer could be that voters instantly
blame the government for everything. The rise of the welfare state has
expanded the role of the state in many domains to the point that citizens
have increased expectations of what the government can and should
address (see Pharr and Putnam, 2000). Since the Second World War, the
expanded role for the state lias gotten itself involved in promoting full
employment, providing adequate public housing, regulating labour o standard, etc. It is logical, then, for citizens to expect the modem state
111 to do something about economic problems. Voters could incorporate an understanding of the Canadian government as rather interventionist in the economy. Tax incentives and subsidies are structured to encourage and discourage particular types of economic activities. Often, policy is directed to particular industries or regions, which in either case indirectly affect the economic consequences of occupations predominant in a region or an industry. But the government does play a role in deciding which occupations deserve encouraging, be it through training programs or tax incentives. Therefore, voters who feel their particular occupation is no longer as valued begin to regard themselves as political outsiders, unable to influence government decisions in their favour. The awareness of such a reality would logically lead voters to see representative institutions as less responsive to their needs, and would logically lead such voters to withdraw from politics altogether or to support more “radical” parties. In either case, sucli voters see their democracy as a failure.
An extension to that line of reasoning leads to a conclusion that voters evaluate economic conditions at the level of an occupational reference group not merely as prosperitv or lack thereof, but as a measure of fairness. If after 10 years one see his or lier orcupation continue to suffer while other occupations hold their own, or even grow, one begins to question whetlier higlier powers bave remained loyal to a some social contract of equitable distribution of costs and benefits, especially given an institutional setting that is characterized by policies
112 0 oriented towards providing generous welfare and
programs ai-id an inter-governmental equalization payment scheme, that
taken as a whole, may lead citizens into believing that the state is set up
to provide some measure of equality. Anv deviation from such a path
would lead the losers into believing that the system is beginning to fail
them.
AÏthough the data presented here does not directly validate such
an argument, there is sufficient evidence to suggest something along
those unes. There is a strong indication that voters in occupations that
have been suffering long-term economic decline are more likely to
behave as if they are clearly very upset at the political system. What is
required now is to explore further these initial Hndings and the extent to
which economic decline can drive a wedge among different groups of
voters, and whether this wedge is understood as the failure of a
democratic system’s promise to deliver governance equitably.
113 o
CHÂPTER 3
ECONOMIc DISPARITIES AND NON-MAINsTREAM VOTING:
THE DISCONTENT 0F WORKING-CLASS VOTER WITH NO
PosT-SEc0NDARY EDucATION
This article is under review at Political Research QuarterÏy.
114 Introduction
Any elected government that presides over economic decline is sure to lose support among the electorate. On this point, standard economic voting researcli is clear. But, if within a context of generalized economic growth (or even in a recession), some groups’ circumstances keep falling over a long period whule others do not fare as badly, members of the disadvantaged group can be expected to question the overail fairness of the political system, a sentiment that can rattie the foundation of a political system. The potential for long-term economic to galvanize opposition towards the entire political system is something economic voting research appears to overlook.
In order to fui this void, a certain demographic group will be analyzed to determine whether its long-term economic decline over a long period of time explains its increased propensitv to support non mainstream parties, and whether that electoral support yields implications for the overali legitimacy of the political system. The objective is to replicate some of the findings reported bv Teixeira and
Rogers (2000), whose work shows how since the early Ï970s a particular demographic group in the United States has seen its economic conditions stagnate. After the post-Second World War hevdays when everyone could expect to work hard and earn a decent living, the economic restructuring in the United States since the earlv 1970s has
had consequences that adverselv affected members of the white
115 working-class that lack post-secondarv education. It is flot to say that
only this group suffered, but. according to Teixeira and Rogers this group
bas seen its economic conditions deteriorate more than most others,
and more importantly, this group bas a high voter turnout rate, making
any discontent politically salient. 1 Therefore, as they began to perceive
that the gains and pains of economic progress are not being equitably
distributed, they began to perceive the political system as unfair,
perhaps focused more on elevating the status of other segments of the
American population, such as minority groups and other “special
interests.” As a result, members of this lagging group emerged as an
angry block of swing voters (i.e., the “forgotten majority”) that bas
benefited the Repubuicans, which bas capitalized on anti-government
sentiments 2
This sort of analysis was applied only to the American electorate,
but the economic changes experienced in the United States since the
1970s is a more global phenomenon. It is therefore useful to examine
whether long-terni economic decline bas had a similar effect on voter
behaviour in other countries. Canada is an obvious choice for a casc
study since its economy is highly integrated with that of the United
States, and likely to be affected by similar economic changes. However,
I Other research shows working-class unemployed American voters, regardless of race, tend to abstain from voting, unless mobilized during election campaigns that raise the salience of economic considerations (Southwell, 1996). 2 One prominent feature associated with the Republican historical gains in the 1994 mid-term was Newt Gingrich’s “Contract with America” campaign theme O and its focus to clean up, or dismantie, government.
116 ..inlike American polities, Canada lacks a racial cleavage. But in Canada,
as in the United States, those in the working-class who lack post
secondary education bore the brunt of economic changes, especially since the recession of the eariy 1980s.
Three fundamental changes have impacted this group: 1) The rise of the modem “knowledge based” economv, which depends more on a highly educated workforce, leaving behind those who lack proper training; 2) the increase in the number of women who pursue post secondary education and the increase in labour force participation rates among women, making the labour market more competitive; and 3) the decline of the indu striai sector, which has been replaced by the growth of the service economv, a sector that has not readilv replaced the high union wages of manufacturing plants. These changes may have contributed to the creation of a new underclass which, since the 1980s, has behaved politically in ways that can be interpreted to reflect their eroding conditions. As the analysis here will show, since the 1980s those in the working-class who lack post-secondary education have been voting increasingly for what is referred to here as “non-mainstream parties,” parties other than the Liberal and Progressive Conservative, and this voting pattemn is hnked to a rejection of the political system.
Others have detected a similar phenomenon in the United States. Third party presidential candidate Ross Perot has attracted American voters who have suffered iong-term economic decline or feel a greater sense of job insecurity (Mughan and Lacy, 2002) and who regard the political
117 system as unresponsive (Southwell and Everest, 1998). But sucli “third
partv” candidates rareiv emerge to mobilize disiiiusioned American
voters, who otherwise opt not to vote at ail.3
In contrast, Canadian voters generally often can choose a non mainstream party, and often, sucli a party can be considered a viable force that can easily distinguish itseif from the mainstream. Since
Confederation in 1867, elections were primariiv a two-way race between the Liberal and Progressive Conservative parties.1 Both parties formed within parliament at the founding of Canada; both parties alternated as government and officiai opposition; and both parties emerged as two political entities most cleariv associated with the Canadian political system.5 But other parties have appeared on the ballot, aside from the prominent “third” parties that rose.
Smaller, iesser known and more obscure parties field candidates and garner votes, but remain for the most part completely overlooked both by the media and by other political observers. This perhaps reflects a reality of voting in the context of Canada’s piurality system. A voter
Ross Perot’s relative success can be attrihuted to his personal wealth, given that money is an important resource in American politics. But monev is flot the onlv obstacle to potential “third parties” in the United States. The dominance of the two-party system may be expiained, at least in part, ta the primarv system, which tends ta draw a diverse arrav cf candidates, many of which are more ideological extreme (see Bradv, et al., 2005). In other words, candidates and voters who would otherwise belong to non-mainstream parties converge around the general ideological umbrella of either of the two mainstream parties. The original Conservative party of Confederation changed its name to the Progressive Conservative in the 1940s. It should be pointed out in 2003 the beleaguered Progressive Conservative party merged with the Alliance Party to form the new Conservative Part. Before then, and especially before the 1993 election, Canadian politics was dominated by the Liberal and PC parties.
118 who wishes to support any of the non-mainstream parties must ponder
the extent to which he or she is prepared to cast a wasted ballot.
Canada’s electoral svstem favours large and centrist parties, almost
guaranteeing other parties no more than a prominent presence in
opposition. Hence, a voter who shifts support from one mainstream
partv to the other expresses a very different sort of message compared to
one who shifts from any mainstream party to one that is non
mainstream. The latter requires far more “connection,” since non
mainstream parties are not normally linked with politics as usual, and
are also not expected to form the government.
As pointed out before, voters have mobilized around new and
emerging parties, usually as a protest against what was perceived as a prevailing injustice. Often, the emergence of a non-mainstream party is based on economic discontent with the two mainstream offerings. The
Social Credit and the forerunner to the New Democratic Party, the Co
Operative Commonwealth Pederation, were formed during the Great
Depression. Other non-mainstream parties appear f0 have risen out of some sort of political grievance. For example, the Reform Party emerged in the Western provinces in the 1980s out of a widespread and growing sense of alienation and a belief that fundamental reforms (e.g., the
“Triple-E” Senate) are needed to rebalance Canada’s parliamentary system, in light of an apparent bias in favour of the two largest, most industrialized and most vote-rich Central provinces: Quebec and
Ontario. The Bloc Québécois, which asserts the independence of
119 Québec, grew out of a sense of dissatisfaction with Canada’s inability to
address constitutional reform to accommodate Quebec’s traditionai
demands for more autonomy. Other, lesser known non-mainstream
parties, such as the Communists, Libertarians, Gi-eens, and so forth,
also promote fundamental changes. Even the NDP runs on a platform of
fundamental reform, be it to the electorai system, the banking system,
etc. But in ail cases, non-mainstream parties promote policy platforms
that address deep grievances by recommending fundamental changes.
And they do this not only as part of their raison d’être, but also to be
noticed. As Landes (2002) points out (sec also Taggart, 1998), the policy
platforms of non-mainstream parties are a stark contrast from the more
pragmatic stands taken by the mainstream parties, and often “attacks
the existing order and suggests the outiines of a new pattern of power
relationships,” (364) (see also Taggart, 1998). They criticize how Canada
is governed, and assert fundamental reform at the sysem level. Also,
non-mainstream parties share in common their extra-pariiamentary
origin, making them effectiveiy outsiders in a politicai game of
established insiders.
Perhaps as a reflection of their more marginal impact, non
mainstream parties have not drawn nearly as much empirical attention
in Canada, at ieast not before the watershed 1993 election. Studies that
seek to explain why voters would support non-mainstream parties are
relatively few and far between. Perhaps the most weli known e examination of this phenomenon in Canada is from Maurice Pinard
120 (1971, 1973), who persuasively points to economic “strains” that have
led to the rise of Social Credit in Quebec. But more recent work suggests
that the increase support for nonmainstream parties stems from
discontent towards, and dealignment from, the two mairstream offerings
(CÏarke and Kornberg, 1993, 1996; Clarke, et al., 2000), a move that
may be driven by anti-state, anti-system, or anti-party sentiments
(Bélanger, 2004; Gidengil, et al., 2001; see also Poguntke, 1996, and
Sartori, 1976). Another obvious feature of the relatively recent success of non-mainstream parties is their regional orientation. As already mentioned, the Reform Party was founded to address a widespread
— sense of Western alienation and pushed for among other things — the reform of Canada’s Senate in order to raise the political influence of
Western Canada, while the Bloc Québécois formed in 1990 after the collapse of the Meech Lake constitutional agreement that would have met the demands of Quebec nationalists. and has since been advocating the sovereignty for Quebec. Clearly, both parties’ appeals are non economic.
However, support for these two non-mainstream parties lias varied from one election to the next, and in the case of Reform (and in its later incarnation as the Alliance Partv) across regions, as well. Ail the while, the relative strength (or weakness) of Western provinces in the
Canadian federation, and Quebec’s relationship with the rest of Canada, has remained relativelv constant. Again, it must be pointed out tliat other non-mainstream parties, such as the New Democratic Partv, as
121 well as more obscure parties such as the Marxist-Leninists and
Christian Heritage, have garnered votes, witli support for ail of these
non-mainstream parties varving across time and space. This variation
cari be attributed, as the research here will show, at least in part, to
economic factors. More precisely, the post-1980 economic decline of
those in the working-class who do not possess post-secondary education
lias elevated their propensity to support non-mainstream parties. Such
a vote is linked to a rejection of the political status quo, and this is
substantiated by looking at this group’s post-1980s drop in its overail
sense of external political efficacy. Since the 1980s, this demographic
groups already cynical attitude towards the politicai system grew worse,
whuÏe their support for non-mainstream parties grew stronger. All of this
suggests that voters who suffer long-term economic decline are prone to
withdraw their support for the entire political system, and gravitate
towards parties that advocate fundamental change. As highuighted in the
next section, the prospect for such a phenomenon appears absent in the
otherwise robust models that link economic conditions to voting
behaviour.
Re-Conceptualizing Economic Voting
Economic voting has produced some of the most robust models in Q voting researcli. Then again, the basic premise of economic voting is not
122 overly counterintuitive. Put simpiy, voters hoid the government responsible for economic conditions; ruling incumbents are more likeiy to be re-eiected whule presiding over good economic times, whiie electoral support tends to decline with economic stagnation. This is the main thrust of the “responsibility hypothesis.”
How this plays itself ouf varies from one institutionai or sociological context to another. For example, voters are more iikeiy to attribute personat economic circumstances to govemment decisions (i.e., the egotropic effect) in institutional contexts where economic conditions are more closely dependent on government intervention, such as in weifare states (Nannestad and Paidam 1994, 1995, 1997). Voters in societies with a more individualistic political culture, such as in the
United States, are not as likely to hold the government responsibie for personal economic circumstances (feidman, 1982, 1985; Sniderman and Brody, 1977). But in almost ail cases, economic voting almost aiways includes an assessment of the national, or sociotropic, economy, whiie personai considerations vieid generaHy weaker effects (Lewis-Beck,
198$).
Virtually ail economic-voting studies focus on short-term economic changes. Sometimes the indicators are real-world statisticai data, such as the change in unemployment rates one year prior to an eiection, or voter perceptions, which are tapped by election-survey questions througli which respondents evaluate short-term economic conditions. Some noteworthy examples of each include Kramer (1971),
123 who linked aggregate-level income to support levels for US. House of
Representative members of the same party as the incumbent president.
Work by fiorina (1978), Kinder and Kiewiet (1979, 1981) are examples 0f
the latter, where survey data are used to illustrate the electoral
consequences of voter perceptions about the economy. Such approaches
produce highly accurate predictions about the electoral chances of an
incumbent vis-à-vis its main opposition, especially in trying to predict
resuits of a US. Presidential election (see Lewis-Beck and Tien, 2004).
But rarely does economic voting research extend beyond the short-term.
Voters are assumed to be myopic, able to recail and react to more recent
economic change, rather than more long-term change (Nannestad and
Paldam, 1994).
A main contention here is that a short-terni focus restricts the
ability to explain what is arguably a more important implication of
economic change: support for the political system as a whole. Whereas
short-term economic changes clearly have an implication for the sitting
incumbent, since a one-year change in something like the
unemployment rate can be tied to government performance and the
competence of its leaders, long-term economic changes lead voters into a
different mmd frame. As will be demonstrated, a long-terni perspective
suggests a vote calculus that is not merely an expression of approval for
or against the current administrators, but a reflection of whether one
regards the political system as legitimate. C
124 It takes time for voters to react along the unes elaborated liere. If a particular group in society suffers long-term economic decÏine, its support for non-mainstream parties mav increase onlv marginally from one vear to the next: observations made over a short period of time may not detect any movement. Yet, there may very well be a movement that suggests increased disapproval of the political system as a whole. But onlv a long-term perspective would be able to observe definitively such subtie patterns.
Standard economic voting research lias littie to say about this possibility, but it is flot as if sucli a possibiuity lias littie theoretical basis. Easton (1975), for example, lias stated tliat long-term decline in the performance of a government can leaU voters to become increasingly disillusioned witli tlie wav they are being governed, pinning tlie blame flot on any one political leader or party, but on the entire political system. Lipset (1959/1963) points out that generalized prosperity is needed in order for a society to maintain political stabulity. In other words, no identifiable group should emerge as an economic loser. This does flot mean that everyone should have the same income or living conditions, but it does suggest that those who experience long-term economic decline will regard tlieir pliglit not as personal misfortune, and flot as the fault of any one party or leader, but as a violation of some social contract. Consequentlv, voters blame the political system for ripping them off. If this process does indeed take place, then such a losing group would be more likely to support political parties that
125 acknowledge the hardship and blame flot any one set of political leaders, but the political system as a wliole.
for sure, scandais and political corruption are bound to lead to similar consequences, but there are more structural reasons that lead voters f0 question the fairness of their political system. And although political reasons may form a large part of that discontent in Canada
(e.g., Western alienation, Quebec separatism, etc.), economics is arguable one of the most important and fundamental drivers. Aside from the fact that healthy economic conditions help voters simply to survive, economic conditions also reveal status. Equality is a fundamental principle in ail democratic societies, and anyone who believes the system has caused them to suffer systemic inequality is not likeiy f0 exhibit
high levels of support for that system. Despite ail efforts to give each citizen the same opportunities f0 advance economically, over time, the allocations of costs and benefits change in a society, with some segments carrying an increasingly heavier burden of the costs. This may be toleratd in the short term, since no one can reasonably expect to be a winner ail the time, given the often rapid and unexpected shifts that characterize advanced industrialized economies. But no one expects to be a ioser ail the time, either. Yet, if this perception emerges, then vofers who belong to a group that suffers persistent economic decline become increasingly aware of their emerging loser status, and consequently, are
expected to begin regarding the political system as no longer able — or
willing — to deliver benefits. In such a situation, the losing segment is
126 bound to react against the entire svstem, and accordingly begin to
express their discontent by voting for non-mainstream parties.
Such phenomenon escapes the notice of conventional economic
voting research. As stated before, the focus on the short-term fails to
capture variations in support for non-mainstream movements, which
tend to be more subtie. This short-term approach is more appropriate to
explain temporary, ephemeral, fluctuations in voter preferences, which
yield measurable implications for mainstream parties, vis-à-vis the
mainstream opposition. But support levels for non-mainstream political
movements also vary over time and across space. In order to analvze
such a phenomenon, voting research must adopt a longer-term
perspective.
Economic changes that unfolded during the last quarter of the
2Oth century offer an opportunity to examine this link between
economic decline and support for non-mainstream politics. The period is
marked bv several deep recessions, high unemployment and major
industrial restructuring, as well as resurgent economic growth
(especially during the second haif of the 1990s). If these changes matter
to voters, then sector-by-sector discrepancies in economic health should
yield political ramifications bevond what standard-economic voting can
detect. o
127 Divergfng Economic Demographic Groups
One could argue that Canada, like manv other industriahzed countries, enjoys consistent gr vth over the long-term and that there is no need to look for economic causes of any simmering legitimacy crisis.
Occasionally there may be a recession, but over the long haul, Canada’s economy continues to expand. Indeed, Canada’s Gross Domestic
Product grew from $556 billion in 1981 to $945 billion in 2000.6 So why is it that amid this prosperous trend, voters have supported non mainstream parties, some of which have enjoved unprecedented success?
Perhaps it is because the long-term growth that characterizes a national economy is experienced differently across sectors. Uver the same long period of time, no two segments of the population are identical; some have enjoyed considerable prosperity, while others have lagged. One group in Canada that appears to have stagnated, at least
more than usual, is the working-class — especiallv those in this group who do not possess post-secondary education. This group has aiways tended to be the most economically worse off, but this demographic group has witnessed a substantial long-term deterioration of its relative situation since the R80s.
b Economic data was retrieved from Statistics Canada CANSIM database, Table No. 379-0017, which is set to 1997 constant Canadian dollars.
128 In order to illustrate these economic trends, unemplovment rates from Statistics Canadas Labour force Suruey were used to generate a time series.7 Two separate groups were defined. Group 1: individuais who lack post-secondary education and are employed in a working-class occupation; and Group 2: ail others. Working class occupations include the following categories of Statistics Canada’s Standard Occupational
Categories: forestry, mining, processing, machining, fabrication, construction, clericaÏ, service and transportation. These are mostly blue collar and the low-paying “white collar” occupations. foremen and other low-level or “shop-level” supervisory occupations are also included; management/ executive and professional occupations (e.g., engineers, accountants, etc.) are net. Agriculture is aise excluded, since it is a unique sector characterized by seasonal fluctuations in activitv, government-support and supply-management systems (e.g., dairy and eggs). It is aise a very small segment cf the workforce.
During the period under study, the proportion of Canadian workforce employed in any cf these occupational groups ranged from a peak of 51 percent in the late 1970s te a low of 44 percent in the mid
1990s. The proportion of the Caiadian workforce that be]ongs te Group
1 ranged from a high of 35 percent in the 1970s te a low of 17 percent in thelate 1990s.
Annuallv aggregated microdata datafiles of the LfS were made available for download from the Sherlock website (http://sherlock.crepuq.qc.ca/), but onlv for survevs 1976 onward.
129 Each groups unemployment rate is plotted in figure 3.1. The
differential between Group 1 and Group 2 is aÏso pÏotted. The higher the
differential, the worse off Group 1 is compared to Group 2. As the graph
clearly shows, s;nce the recession in early 1980s, Group 1 Sas suffered
relatively higlier unemployment rates, a trend that continued to get
worse until the mid-1990s, only to level off slightly afterwards. But the
differential remained above its pre-1980 levels. Unemployment rates and
the associated differential une reflect three-vear moving averages.
Figure 3.1: Unemployment rates
25 CI, C) 20
C) 15 r’
C 10 2 G C) 5 1
0 - cn 3 O (N c’) - L!) çD N OD O O — c c’D zi- o N
,- - - .- - - - ,- - -.. - ,- - - - .-1 - 4 -
Une m plovment rate: Economicallv disadvantaged group Unemplovment rate: Ail others Differential x lO-vear trend
Source: Statistics Canada, Labour force Survey o
130 The series entitled “lO-vear trend” vas generared using Excel’s
TREND function, which uses OLS to produce a regression coefficient.
For each vear from 1985 to 1998, a trend coefficient was calculated
using the previous 10 vears as the independent variable and the
unemployment differentials of those 10 years as the dependent variable.
For example, the “trend” at 1997 was 4.3, whicli is the siope coefficient
for the regression model based on unempÏoyment differentials between
1988 and 1997.8 Positive values suggest worsening conditions. As the
trend line clearly shows, the unemployment differential between Group
1 and Group 2 grew increasingly pronounced since the mid-19$Os.
Studies conducted by Statistics Canada bear out this demographic group’s economic decline. Crompton and Vickers (2000), for instance, report that since the 1980s, goods-producing and manufacturing jobs restructured as service-oriented industries expanded. Also, economies in the developing world were increasinglv dominant in labour-intensive industries, leading to many plant closures in Canada and in other highlv advanced industries. Older men in particular were more likely to face permanent layoff, especially those without proper education. Younger workers also suffered, especïally since the 1990s. But in general, members of the working-class who lacked sufficient education have faced a long-term decline. At the same time, women were participating more in the labour market, arid thev
R Since the available LfS microdata starts from 1976, the earliest long-term trend can onlv 5e generated for the 1976-1985 period.
131 were also acquiring more and more education, gaining ground in the
labour market (see also Heisz et al., 2002, and Picot and Heisz, 2000).
It is clear that the increased demand for more educated workers,
and the relocation of labour-intensive industrial processes to the
developing world has hit liard the working-class, especially those who lack post-secondary education. This particular segment, already prone to suffer more economically, suffered the rnost amid the post-1980s economic restructuring. However, economic policy may also have played a role. Crompton and Vickers’s study (2000) also suggests that shifts in labour market dvnamics were a resuit of an anti-inflationarv policv that was aimed at the stagflationary economv of tlie late 1970s. Such a policy slows down the economv and, consequentlv, increases unemployment.
If, indeed, economic restructuring can be attributed to policies that were initiated by the Liberals in the 1970s, continued by the Progressive
Conservatives in the 19$Os-1990s, and continued once again by the
Liberaïs since 1993, then it should corne as no surprise to sec voters from that most severely affected group react against both of the mainstream parties, and even against mainstream poÏitics altogether.
Data and Methods
Canadian Election Study surveys from 1974 to 1997 were gathered to determine whetlier Group 1 votes any differently from Group
132 2. A dummy variable was constructed whereby a score of I is assigned
to respondents who are employed in a working-class occupation1° and
have no post-secondary education, and a score of O to ail others.
Respondents’ reported voting behaviour was recoded into a variable with
the following three categories: 1) vote for incumbent; 2) vote for
mainstream opposition; and 3) vote for non-mainstream.
Since, as shown in Figure 3.1, the difference between the
economic circumstances of Group 1 and Group 2 only began to diverge
sharply since the early 1980s, election surveys were compiled into two
separate datasets. The first dataset encompasses the resuits of the 1974
and 1979 CES surveys, and the second dataset encompassing the 1984
to 1997 surveys.11 This arrangement permits a before-and-after analysis
of voting behaviour.
Other variables were included as control variables, mostlv to
account for some of the main cleavages that characterize Canadian
° Data from the 1974 and 1979 Canadian Election Studies were made availabie by ICPSR. Data were originally coliected by Harold Clarke, Jane Jenson, Lawrence Leduc, and Jon Pammett. The 1984 to 1997 data were provided by the Institute for Social Research, York University. The 1984 data were collected by R.D. Lambert, S.D. Brown, J.E. Curtis, B.J. Kay and J.M. Wilson; the 1988 data were coiiected by Richard Jolinston, André Biais, Henry E. Brady and Jean Crête; the 1993 data were coiiected by Richard Johnston, André Biais, Henry Brady, Elisabeth Gidengil, and Neii Nevitte; the 1997 data were collected bv André Biais, Ehsabeth Gidengii, Richard Nadeau and Neii Nevitte. Data from the 1984-97 Canadian Eiection Study were funded by the Social Sciences and Humanities Research Council of Canada (SSHRC). Neither the ICPSR, SSHRC, nor the original collectors of the data bear any responsibility for the analyses or interpretation presented here. t) The CES surveys track the respondents’ occupations as per the Statistics Canada categories. 1 The 1980 and 2000 CES were excluded because they do not track Q occupation.
133 electoral politics (see Biais, et al., 2002). These variables are: region,12
unionization. religion, 4 language, gender 16 and age. Dummv
variables were also included to account for particular election vears,
namely, 1974, 1984 1988 and 1993.
Once the variables were assembled and compiled into the two
separate datasets, multinomial logistic regression modeis were
generated to test the following hypothesis:
H3. 1: Working-cÏuss voters who lack post-secondary educatton are
more Ïikely to vote non-mainstream compared to alt other voters in
elections held after the 1 980s compared to elections prior to the
1980s.
The dependent variable is vote choice, with “vote for incumbent” set as
the base category. The Liberai party was the incumbent in 1974, 1979,
1984 and 1997, while the Progressive Conservative party was the
incumbent in 1988 and 1993.
2 Three regional dummy variables were constructed: Atiantic (Newfoundland, Nova Scotia, New Brunswick, and Prince Edward Isiand); Quebec; and West (Manitoba, Saskatchewan, Alberta and British Columbia). Respondents who live in one of the three regions are given a score of 1 on the appropriate dummv variable, O otherwise. Ontario vas seiected as the reference group. Respondents from anv of the northern territories were exciuded due to smaH sampie sizes. 3 A score of 1 is assigned ta n respondent who belongs ta a union. 4 Cathalics are assigned n 1, ail others, including those who are flot afflhiated with any other religion, are assigned a 0. 15 Francophones are assigned n J, ail others a 0. Q 6 Maie=l, female=0.
134 Table 3.la shows that in the elections heid from 1984 to 1997, a
respondent from Group 1 is more likely to vote for non-mainstream
parties, compared to supporting the incumbent party. 17 Belonging to
Group 1 has no bearing on voting for the mainstream opposition party
compared to voting for the incumbent. In other words, voters do not
appear to differentiate the two mainstream parties, regardless of
economic conditions. But having suffered long-term economic decline
raises the probability of voting non-mainstream over the incumbent. It
must be pointed out that this effect is flot the strongest (the highest
coefficient is the “Western Provinces” variable, reflecting Canada’s strong
regional cleavage), but it is significant even while controlling for other
factors.
Analysis was repeated with the 1974 to 1979 dataset, and yielded
expected resuits (Table 3.lb). As per the first hypothesis, belonging to
Group 1 lias no bearing on the probability of voting non-mainstream
compared to the incumbent for elections held before the 1980s. The
“Economically declining demographic group” variable fails to achieve
statistical significance. Again, the strongest factor is region.
‘ See Appendix 3 for a complete and detailed description of ail variables used C in this article.
135
Q
p<.001
p<.00l p<.001
p<.001
p<.Ol
p<.001
p<.00l
p<.00l
p<.05
p<.05
Err.
.133
.083 .111
.094
.070
.066
.002
.136
.084 .081
.139
.124
.073
elections
Std.
Non-mainstream
.067
.874
.516
.961 .477
.152
-.034
-.151
Coef. -.401 -.902
Canadian
-.007
-.315
-.297
1997
to
1984
p<.001
p<.001
p<.001 p<.Ol
p<.Ol
p<.Ol
choice,
opposition
Em
vote
.134
.098
.071 .121
.095
.002
.076 .123 .064
.101 .081
.124
.069
of
Std.
party.
Mainstream
.003
.027 .028
.023
.184
.296
.317
Coef.
-.169
-.084
-.027
-.952
1.934
1.686
estimates
incumbent
is
logit
.102
6549
-6445.60
category
dedining Provinces
group Canada
Multinomial
base
dummies
la:
3.
The
Quebec
Atiantic
Western
likelffiood:
1984
1988 1993
Constant Pseudo-R2: Log Note:
N:
Union
demographic Religion Gender
Economically Regional
Language
Age
Table C
o
p<.Ol
p<.lO
p<.001
p<.001
p<.00l
p<.lO p<.O5
p<.05
Err.
.165
.284
.142
.221 .142
.004
.235 .171
.222 .178
.139
elections
Std.
Non-mainstream
.508
.607
.678
-.543
-.031
-.005
Canadian
-.393
Coef.
-.100 -.016
-.388
-.532
1979
to
1974
p<.00l
p<.05
p<.05
p<.001 p<.00l
p<.00l
p.<.lO
choice,
opposition
Err.
vote
.135
.231
.003 .130
.210 .119
.141
.216
.155 .146
.115
of
Std.
party.
Mainstream
.005
.071
.217
.580
.666
-.296
Coef.
-.190
-.286
-.462
-.900 -.886
estimates
incumbent
is
logit
1.08
.104
1827
-170
categoiy
decining
Provinces
Canada
group
Multinomial
base
dummies
lb:
3.
The
Quebec
Western
Atiantic
likelihood:
1979
Note:
Log
Constant N:
Gender
Union
Language Pseudo-R2:
demographic Regional Religion
Economically
Age Table The generalization that can be drawn from these two sets of
resuits is that voters in the economically weake’ Group 1 did not vote
any differently from those who belong to Group 2 during the more
“equitable” period before the pivotai 1980s, but when economic
conditions became less “equitable,” voters in the worse off Group 1
began to vote in a manner that is clearly more distinguishable and more
suggestive of their growing discontent.
The next step examines the link between economic decline and
vote choice through the mediating effects of external pobtical efficacy,
which measures the extent to which a respondent feels that the political
system is responsive to his or her needs. Political efficacy was originaliy
set up out of a more general five-item scale’8 (Campbell, et ai., 1954).
Subsequent research (Baicli, 1974; Craig and Maggiotto, 1982; Lane,
1959) differentiates internai efficacy, which measures the extent to
which a respondent feels personally capable of influencing the political
system, and external efficacy, which bas already been defined.
Efficacy measures have been used in seminal research in political
culture to differentiate societies with more participant populations
(Almond and Verba, 1963). Despite “broad brush” descriptions about an
entire society’s pobtical culture, efficacy levels can vary across sectors
18 The original survey items are: j) “I don’t think public officiais care much what people like me think;” ii) “The way people vote is the main thing that decides how things are run in this country;” iii) “Voting is the only way that people like me can have any say about how the government runs things;” iv) “Peopie like me don’t have any say about what the government does;” and y) “Sometimes politics and government seem so complicated that a person like me can’t really O undersfand what’s going on.”
138 within the same society. For example, one’s efficacy lias been linked to
socioeconomic status (Wu, 2003). In Canada, Gidengil (1990) finds lower
levels of efficacy among Canadians living in more “peripheral” economic
regions, compared to the higher scores of those who reside in “central”
economic regions. This suggests a link between economic circumstances
and attitudes towards the political system.
As a concept related to political culture, efficacy is a durable
personality trait that develops through socialization and is unlikely to
fluctuate easily and quickly (Easton and Dennis, 1967; Iyengar, 1980).
But, as mentioned earlier, even Easton (1975) recognizes the possibility
that long-term decline in conditions can lead people to hold more critical
views of their political system. Therefore, those who suffer Iong-term
economic decline, such as respondents who belong to Group 1, are
expected to manifest lower levels of external political efficacy compared
to those, such as in Group 2, who have not fared as badly. In addition,
the durability of attitude traits means that efficacy is not likely to
change over the short-term. Therefore, if there are any changes to one’s
level of efficacy, it must be examined over the longterm.
Since Group 1 ‘s economic decline compared to Group 2 began in
the 1980s, an “efficacy gap” is expected to emerge, and in turn, this
“efficacy gap” is expected to explain the link between long-term
economic circumstances and the propensity to vote non-mainstream.
These expectations can lie summarized with the following hypothesis: o
139 H3. 2: Woi*ing-class voters witli no post-secondary crin cation have
lower Ïevels ofexternaÏpoÏitical efficacy than other respondents.
I-133: PohticaÏ efficacy is o mediating factor that expÏains why
working-class voters with no post-secondary education are more
likely to vote non-mainstream afier the 1980s.
External political efficacy was computed using two items found in the eÏection surveys (except the 198$ CES), namely: j) “J don’t think that the government cares much what people like me think,” and ii)
“People hke me don’t have any say about what the government does.”10
The scoring was set to a O to 1 range, where O indicates disagreement with the statement and 1 indicates agreement. In such a scoring scheme, high values suggest Ïower levels of efficacy. Figure 3.2 shows trends for each demographic group. Three main observations can be drawn from that chart. First, ail respondents show a general decline in external political efficacy. Second, as expected, respondents in Group 1
(solid une) are consistently less efficacious (score lower on the political efficacv index) than respondents in Group 2. 20 Third, and most importantly, the gap between the two groups widened since the mid
1970s, peaked in 1984, but neyer recovered to pre-1980s levels. Clearly,
‘° The external efficacy index yields a Cronbach’s alpha of .6601 for the 1974- 1979 dataset, and .6089 for the 1984-1997 dataset. 20 Group l’s lower sense of political efficacy compared to Group 2’s during the 1984-1997 is confirmed through an independent samples t-test (t=-13.575, df=5944, p 140 O it appears that nembers of the working-class who do not pussess secondary credentials have become increasingÏy disillusioned with the political system. Figure 3.2: Polïtical efficacy, 1974-1997 .50 ...--..-. .00 ZZZZzzjj: .20 - 0 10 Q .00 - -- -.14 1974 1979 1984 1993 1997 EconomicaHv dcctining demographic group Allothers Differential Source: Canadian Election Studies, 1974-1997 The next step is to determine whether this trend is tied to non mainstream voting. Two steps are involved. First, ordinarv least squares regression analysis was conducted to determine the extent to which membership in the generally more disadvantaged Group 1 explains sense of political inefficacy. Resuits in Table 3.2 show that, overali, o 141 belonging to Group 1 vields a lower sense of political efficacy.2 This is true for both the pre-1980s and the post-19$Os periods, aithougli the effect is higher in the latter time period, with a coefficient of - .072 for the 1974-1979 period compared to -.105 for 1984-1997. Furtliermore, the variable “Economically declining demographic group” is the most influential, with a beta score of -.170 for the 1974-1979 period, increasing to -.177 for the 1984-1997 period, whule age ranks in second place. No other variable appears to yield a greater weiglit on political efficacy — not even region. In the second step, the models in Table 3.la were re-run witli the politically efficacy variable (see Table 3.3). As expected, membership in Group 1 loses its signfficance in explaining vote probabilities when external political efficacy is included in the model. This suggests that political efficacy functions as an intervening variable, mediating the effect of long-term economic decline and the propensitv f0 vote non mainstream. furthermore, political efficacy emerges as a major factor, ranking a close second behind the “Western Provinces” dummy. A simulation (sec Figure 3.3) was generated to illustrate the relationship between the key variables.22 Belonging to Group 1 raises 21 The OLS analysis does flot include respondents from the 1988 wave because efficacy items were not employed during that election study. As a resuit, the 1988 dummy was dropped. 22 figure 3.3 was generated with Stata using Clarify, a macro developed by Tomz, Wittenberg & King (2003), and available on http://gking.harvard.edu/ stats.shtml. (See also King, Tomz & Wittenberg, 2000). Its purpose is to demonstrate graphically the distribution of statistical models using simulation techniques. The charts show the relationship between the economic variable (whether one belongs to the economically disadvantaged group) and flic vote 142 0 the probabiÏity of voting non-mainstre. Furthermore, Group l’s average politicai external efficacy score of .32 is 10 points behind Group 2’s average of .42. The general pattern here is fairly clear: Long-term economic decline leads voters to hold more negative attitudes about the politicai system, and in turn, this increases their chances of voting for a non-mainstream party. 23 variable, while setting ail other variables (the control variables) to their mean values. 23 Analysis was repeated with different versions of non-mainstream. If NDP voters are exciuded (under the assumption that the NDP is not a true non mainstream party), then the reiationship between iong-term economic decline and non-mainstream voting fails to attain significance. This is probably due to the fact that the exclusion causes the proportion of respondents who daim to have supported a non-mainstream party to drop from about a third to about a quarter. In another version, fringe and more marginai parties are excluded, but this has not produced different resuits. This discussion is taken up further lin the Conclusion on page 162. 143 C - p<.001 p<.O5 p<.001 p<.001 p<.001 p<.Ol p<.Ol p<.00l p<.Ol p<.00l Err. .011 .008 .014 .007 .008 .016 .000 .009 .009 .016 .012 .008 Std. 1984-1997 - .055 .269 .042 .468 .025 .031 5527 .050 -.008 -.003 Coef. -.026 -.002 -.030 -.048 -.105 efftcacy, lower. p<.00l p<.05 p<.Ol p<.00l p<.05 p<.00l or political Err. .026 .015 p<.O5 .014 .021 .013 .000 .016 .017 .019 .023 .013 Std. at 1974-1979 efficacy external of - - .065 .205 .013 .013 1066 .596 .032 -.008 Coef. -.024 -.025 -.002 -.056 -.053 -.072 political significant analysis elections External statistically regression Provinces Canadian declining Canada group variables: are variable: OLS dummies 1997 3.2: to Quebec Atlantic Western values 1984 1979 1993 1974 Bold Constant Gender N: Union SEE: Language Religion demographic Economically Regional Independent Adjusted-R2: Dependent Age Table o .001 p<.Ol p<.O5 p<.001 p<.001 p<.Ol p<.001 p<.001 p<.001 p< Err. .185 .098 .131 .003 .094 .104 .087 .189 .186 .110 .157 .099 .161 elections Std. Non-mainstream .051 .083 .146 .426 .505 .755 -.291 -.011 Coef. - -.009 -.637 1.014 1.119 -1.032 Canadian 1997 to 1984 p<.001 p<.00l p<.001 p<.O5 p<.001 p<.001 p<.001 p<.lO p choice, opposition Err. .102 .181 .096 .138 .003 .176 .177 .085 .100 .107 .155 .136 .094 vote Std. of party. Mainstream 04$ .077 .108 .206 .342 .397 .512 -.454 -.001 -. Coef. -.297 -.837 1.734 1.968 estimates incumbent is logit .114 402$ -3869.25 efficacy categrny declimng Provinces Canada group Multinomial base dummies Political 3.3: The Quebec Western Atlantic likelffiood: 1993 1984 Gender Log Religion Constant Pseudo-R2: Regional Union N: Note: demographic External Economically Language Age Table o Figure 3.3: Simulation of non-mainstream voting and political efficacy .45 .40 .35 .30 .25 .20 - Economically Ail others disadvantage cl group Prob. of voting non-mainstream —+—- Level of poli tical efflcacv Conclusïon The analysis reported here leads to three key conclusions. First, there is a link between long-term economic decline and a propensity to vote for non-mainstream parties. More precisely, long-term economic decline affects how voters evaluate the entire political system. This expresses itself not oniy in lower external political efficacy, but more c&crete1v as a vote against those political parties that are associated with the political system. Clearly, the history of Canada as dominated by either Liberal or Progressive Conservative governments leaves voters with no doubt that a vote for any of these parties endorses the status 146 quo, while a vote for a non-mainstream partv sends an entirelv different sort of message. Second, research could benefit by expanding the perspective with which we study the link between economic conditions and the vote. The standard approacli of examining only the short-term, be it through survey questions or actual real-world conditions, is too narrow. The extent to which such a short-sighted perspective can inform electoral outcomes cannot extend much beyond accounting for the popularity of the incumbent government. A long-term perspective reveals something far more critical, with implications that are not restricted simply to the incumbent or the mainstream opposition, but to the entire political system. Third, and this probably deserves more attention, voters appear sensitive to what may be perceived as changes to a “social contract.” When the benefits and burdens of society shift from one sector to another, the losers appear to regard this shift nof as a random act for which no one can be blamed. The economic shifts that occurred during the 1980s clearly adversely affected some Canadians more than others, and they behaved politically accordingly. But it may flot be economic decline, per se, that vields potentially destabilizing political consequences, but a sense of being “shafted.” At no point in the data series was Group 1 ever in an economically superior position. But being economically weaker is not what drove some voters in Group 1 to vote non-mainstream; it was the widenin.g gap between it and Group 2. If 147 equalitv, or egalitarianism, is a Canadian value, then the economic inequality that grew wider since the 1980s mav have left some voters feeling unfairly punished by the political system, leading many of them to support non-mainstream parties that are not tainted bv the very sanie svstem. 148 CONCLUSION The main point of the research presented here is that at the margins of a healthy democracy, non-mainstream parties garner support from voters, support that must be understood as more than simply a choice like any other. A voter who chooses a non-mainstream party is expressing a different sort of message than one who supports any of the mainstream offerings. In particular, those who support non mainstream parties tend to be motivated by a higher level of discontent directed towards the entire political system, and their choice to support a non-mainstream party reflects a generalized rejection for politics as usual. An examination of voter behaviour in this light brings together two different subfields of political science, namely political sociology and its emphasis on legitimacy, and political behaviour and its emphasis on choice. The findings presented here show that the legitimacy of the political system can be measured through individual vote choices, which in turn can be explained through structural economic changes. The particular condition examined here is the economy, particularly long-term economic decline, which yields implications that are quite different than what standard economic-voting models could reveal. By looking at economic voting through a long-term perspective, economic changes are seen to weigh not on a voter’s evaluation of the incumbent party’s competence, but instead on the overail sense of 149 attachment and legitimacy for the whoÏe political system. In standard economic voting, voters upset witli the incumbent would normally shift support towards the mainstream alternative. The findings here suggest that long-term economic decline hads to no discernable shift of electoral support toward the mainstream opposition. Instead, long-term economic decline benefits non-mainstream parties. This general hypothesis is confirmed by the flndings contained in each of the three articles, each of which adopts a different approach. Chapter 1 followed an aggregate approach. Provincial-level federal election resuits from 1979 to 2000 were pooled together and related to long-term and short-term economic data. The results show that whule support for the incumbent is explained by short-term economic conditions, support for non-mainstream parties is not. lnstead, non mainstream voting is explained by long-term economic changes. Chapters 2 and 3 take these resuits further, by revealing that attitude mediates the link between long—term economic decline and non mainstream voting. Each of the two perspectives reveals a pattern whereby experience with long-term economic decline Ieads one to feel less efficacious, which in turn draws one doser to non-mainstream political parties. In Chapter 2, that experience was feit at the occupational level. Those employed in occupations that were in long term decline were more likely to manifest lower levels of external political efficacy and to support non-mainstream parties, compared to those employed in occupations that did not suffer long-term decline. 150 Chapter 3 showed that those employed in working-class occupations and vho also lacked post-secondarv education experienced an economic decline that was more marked and more severe than what others experienced during the economically volatile period since 1980.’ Members of this demographic group saw their already lower levels of external political efficacy drop further and consequently they voted more for non-mainstream parties. With these flndings, two general conclusions can be drawn. first, long-term economic factors must be quaÏitatively distinguished from short-term economic factors. Research into voting behaviour does flot alwavs make this distinction. And second, long-term economic change is a consistently significant factor in explaining voter support for non mainstream parties. The significance of long-term economic conditions remains resilient even when other, perhaps more “important” political factors are taken into account. Also, the effect appears to bear its weight more heavily on attitudes and orientations towards the political system, a phenomenon that may take a long-time to develop, but an effect that may remain relatively durable. Note that the occupations deflned as “declining” in Article 2 are flot aiways the “working-class” occupations that form part of the definition of the declining socio-demographic group in Article 3. 151 1- Long-Term vs. Short-Term: A Qualitative Distinction One of the mort elaborate voting models, the multi-stage voting model (Biais, et al., 2002; Jackson and Jackson, 2006; Miller and Shanks, 1996), includes “economic evaluations” or “economic perceptions” as just one single component of an otherwise long list of factors that go into the vote calculus (see Figure C. 1). The model distinguishes between the more remote, more sociologicai, and more long-term factors, such as socio-demographic factors (e.g., religion, gender), from the more proximate, individual and short-term factors, such as evaluation of the party leaders. Long-term factors generally explain voter alignments. For example, Catholics tend to support Liberals (see Biais, 2005). Short term considerations are used to explain brief deviations from otherwise well established partisan alignments. In this model, economics is seen as short-term. Hence, Liberal identifiers may withdraw support from the Liberal party if it presided over an economic recession. Indeed, this might partly explain the Liberal election loss in 1984 and the Progressive Conservative loss in 1993. This pattern appears vaiidated in other studies, in particular those conducted in the United States, upon which this model is based (Miller and Shanks, 1996). G 152 (2 C,) C C cl) o 153 Paradoxically, the study upon which this newer version of the muiti-stage model is based (Biais, et al., 2002) argues that economic perceptions, and flot real-worid economic conditions. explain vote choice. Proof of this is provided by the divergence between a decime in actual unemployment from 1993 to 1997 eiection and perceptions among the electorate that unemployment had actuaiiy gotten worse (Nadeau et al., 2000). This suggests that voters are unabie to evaluate accuratelv short-term economic conditions. If Nannestad and Paldam (1994) are correct that voters are unabie to evaluate economic conditions in the long-term, then the flndings reported by Biais et al. (2002) challenge the notion that voters are even able to get the short term right. Perhaps — and this might be a fruitful avenue for future research — economic perceptions of the short-term are based on actuai experience over a longer period of time and in turn may be mixed up with some other short-term considerations. But the resuits reported in this research project suggest that voters are abie to respond accurately to the short-term. As reported in Chapter 1, support for the incumbent varies with short-term term economic changes. More importantly, the long-term most certainiy yields a predictable consequence, and in ways that are not fuiiy compatible with this muiti-stage model, or with the “voter mvopia” hypothesis. Economic conditions looked at on a iong-term dimension appear to expiain the extent to which voters align or de-align from the mainstream parties. The flndings do flot suggest a new voter alignment 154 along any of the non-mainstream parties, but at the very least, long term economic decline does flot bode well for any of the mainstream parties, leading to an electoral field that includes a broader sense of voter discontent that is aimed at the political system as a whole. This suggests a need for the multi-stage model to separate economic evaluations mto short- and long-term evaluations, or, more accurately, to retain short-term economic perceptions as it is (or even to replace it with “short-term conditions”) and include long-term changes to economic conditions as a new entry. Furthermore, the relationship between long-term economic decline and the propenslty to vote non-mainstream appears mediated by a psychological dimension, politicai efficacv. Therefore, this new stage necessarily precedes values and beliefs. figure C.2 shows a suggested modified version of the multi-stage model. C 155 I I Choice Vote Considerations Individual, Short-term Evaluaions Strategic 1996. I I Govemment Leader of Shanks, Long-tenn Sociological, I and Opinions ______ Evaluauons Miler Issue Perceptions 2006; Economic Jackson, and model I and Party Beliefs vote Identification Values Jackson Economic Underlying 2002; mufti-stage Changes ai., et Long-Term Modffled Biais, C.2: Socio-Demographic Source: figure L I This is not to say that the original multi-stage model is incorrect, but that the vote-choice calculus is a more complex process, and that economics functions as a far more exogenous and antecedent factor than previouslv assumed. Also, if economics plays on both the short and the Iong-term, then economics as a whole emerges as far more central to the vote calculus, and notjust another component. 2- Resiliency of Resuits The second general conclusion flows from the consistency cf the resuits in confirming the general hypothesis, that is, long-term economic decline tends to lead one to vote non-mainstream. No matter liow this hypothesis is tested, the evidence points to the same conclusion. The three different perspectives pursued liere produce solid resuits that confirm the relevance of long-term economic conditions in explaining non-mainstream voting, even when controlling for other variables. However, one potential objection is the relatively weak effect that long-term economic decline appears to yield on the vote. Other variables are sometimes more able to explain non-mainstream voting (see Table CJ). o 157 ______ Table C. 1: Factors that influence non-mainstream voting Article 3- Demographic 1-Province 2-Occupation Group Long-term Among the Among the Second-most economic weakest of the weakest of the • influential factor decline significant factors significant factors Strong factor Strongest factor Strong factor • (esp. living in the Region (esp., living in the (esp., living in the Wst and in West) West) Quebec) —--—t.----- — - — - 198$ (negative) Election year - - 1993 d 1993 Region in particular emerged as a very important factor. In fact, the regional dummy variables are among the strongest predictor of flon mainstream voting. Living in the West, especiaÏly, has a considerable impact on voting non-mainstream. But the regional variables did not negate the effect of economic conditions. Even when accounting for region, which represents many of the key political cleavages that are well known to characterize Canadian governance (sec, for example, Smiley, 1971), long-term economic decline remains significant. A negative regional factor is living in the Atiantic provinces. Voters in this region appear to support mainstream parties fairly consistently. Although there is some suggestion that even there long term economic decline takes its toïl on mainstream voting (sec Figure 1.3c), and almost invariably, non-mainstream support is inverselv 158 related to living in what is arguably Canadas most economically depressed region. What might explain these regional variations? The possibility for such diverse responses to the same stimulus is consistent with an understanding of Canada as divided along key dimensions outlined by Smiley (1971), each of which is more salient in some parts of Canada than in others. For example, the English—french cleavage is arguably more salient in Quebec than anywhere else in Canada, while the East West cleavage seems more salient in Western Canada. But these cleavages are not simply differences of opinion. They reftect a perception by one group that they are disadvantaged, that they are, in effect, losers in a system that is biased. When such a group suffers economic decline, its response may be rooted in light of these cleavages. In sum, the formation of political movements in Canada must be grounded in a regional context, even though the causes may be similar across the country. But this level of heterogeneity fails outside the confines of this project, but it would certainly add a much needed level of snphistication to the understanding of Canadian political behaviour. Therefore, if cleavages vary in their salience from one region to the next, the manner in which voters express themselves will varv. In Quebec, discontent against the entire political system may express itself through support for sovereignty (as was seen afLr the sponsorship scandai broke); in Atlantic Canada, resentment mav express itself through withdrawal and disengagement; in Western Canada, discontent 159 tends to fuel the growth of populisrn: in Ontario, cliscontent tends to fuel support for left-Ieaning parties, such as the NDP (see Henderson, 2004; Simeon and Elkins, 1974). A similar set of factors may explam why non—mainstream parties, particuÏarly the populist and radical riglit. have grown in Europe. Although the rise of sucli parties is not unique to countries in economic decline, there is some suggestion that support within any country seems to be higher among those most likely to have suffered economic decline. For example, support for the non-mainstream right is generally higher among males, whites, members of the lower socio-economic class, those who lack advanced education, (Fieschi and Heywood, 2004), and during times of rising unemployment (Jackman and Volpert, 1996; Givens, 2002). Golder (2003) highlights an interaction between unemployment and immigration. In other words, economic decline and pre-existing orientations (i.e., political culture, attitudes) may interact to explain what encourages or constrains the emergence of this particular brand of right-wing non-mainstream parties. Whether similar factors expÏain why the left was more a prominent non-mainstream movement in a previous time in Europe is something that fails outside the confines of this project, but it is curious that the non-mainstream space typically occupied by the left is not more dominated oy the riglit. Aside from region, support for non-mainstream voting is also affected by factors unique to particular election years, vindicating those who believe election campaigns do indeed have an effect. As shown in 160 Table C.1, the 1993 election shows up as a particularly good year for non-mainsLream voting. while the negative coefficient for the 1988 election dummy (see Table 3.la) reflects the widespread belief of that particular election as being essentially a referendum on the Canada-U.S. Pree Trade Agreement, with voters dividing themselves accordinglv between the reigning Progressive Conservatives and the anti-free trade Liberals (see Johnston, et al., 1992). Certainly, this suggests a need to look beyond just economics as the cause of different electoral patterns. Non-mainstream voting and voter discontent are not only and not aiways based on economic factors, and the research presented here in no way suggests that other, non economic, factors should be ignored. But what is remarkable in ail these results is that even when accounting for particular election vears, region and other factors that characterize the main cleavages of Canadian politics, and factors that clearly are strong and robust, long-term economic decline nonetheless remains relevant. furthermore, economic decline appears to weigh heavily on political efficacy. In both Articles 2 and 3, long-term economic decline emerged as the strongest factor that explains external poiitical efficacy. As shown in Table C.2, region, particular election years, age, and other demographic characteristics appear much weaker.2 Long-term economic decline, whether conceptualized as a decline in occupational growth or a 2 The 1993 election vear ranks a close second in Article 2. 161 D decline of a particular socio-economic demographic group, is strongest factor that brings clown political efficacy. If voters are becoming increasingly cynical and cranky about mainstream politics, and if this troubling trend yields a potential threat to political stability in Canada, then one needs to consider long-term economic decline as a key dimension in electoral poÏitics. Table C.2: factors affecting political efficacy: Comparing beta scores from Articles 2 and 3 Dependent variable: External political efficacy Beta scores Independent variables: Article 2 Article 3 Long-term economic decline Occupation (Article 2) .175 <.001 -.177 p<.00l Demographic group (Article 3) Regional dummies Atlantic Canada -.030 -.058 p<.001 Quebec .056 p<.lO .076 p<.Ol Western Provinces -.051 p<.lO -.052 p<.Ol Religion -.018 -.046 p<.Ol Language - .025 - .005 Age -.089 p<.00l -.087 p<.00l Gender .030 .055 p<.00l Union .032 -.0 13 1984 .075 p<.001 1993 -.156 p<.001 .033 p<.O5 Another potential objection concerns the conceptualization of “non-mainstream.” This category has been defined as any party other 162 than the Liberal or Conservative partv. However, whether the NDP reallv qualifies as non-mainstream should be further examined, given that the NDP bas occupied office at the provincial level and that it bas supported various Liberal minority governments. But when this partv is excluded from analysis, resuits are inconsistent. The models in Article Y are replicated, but the standard error of the estimate grows considerably. In Article 3, exciuding the NDP renders insignificant the relationship between long-term economic decline and non-mainstream voting. Exclu ding the NDP does not appear to have negatively impacted resuits in Article 2. Ail this suggests that it might be prudent to continue regarding the NDP as a non-mainstream party, at least at the federal level. Also, bv exciuding the NDP, the proportion of CES respondents who daim to have supported a non-mainstream party drops considerablv, from about a third to about a quarter. This leads to a bias, whereby analysis of non-mainstream voting would be overwhelmed by support for the Reform/Alliance and Bloc Quebecois parties. Perhaps the NDP’s policy views more directly appeal to economic discontent than the other two parties, a point that merits further exploration. Similarlv, it is worthv to explore whether the results obtained in ail three articles would repeat if analysis excludes marginal parties, those with less than a reasonable hope to win even one seat. But excluding such parties leads to another set of inconsistencies. Resuits in Article 1 and 3 are substantially the same, but in Article 2, the relationship between long-term economic decline and non-mainstream 163 G votmg fails to attam sigmficance. In sum, tlie Droad definition 0f fl0- mamstream here lias some ment. Now, why 1ongterm economic decline has a heavv impact on attitude, but a relatively light impact on votmg, is something that remains unanswered. One interpretation points to institutional constraints Voters who suffer long-term economic decime are more prone to vote non-mainstream, but the plurality electoral system may cause many such voters to abandon their first non-mainstream choice and instead opt for a more winnable mainstream second choice The possibility for such a dynamic is examined bnefly in the next section. 3- Potential Objections and Qualifications The perspective presented here is rather unique. Rarely do researchers in economic voting employ a long-term perspective, and rarely does voting researcli rnterpret resuits as an expression for or against the legitimacy of the entire political system. But in a way the overail results seem rather unsurprising. It makes sense for Yong-term 1osers to blame the system for their declining conditions and eventually withdraw support for that svstem. But despite the intuitive plausibility of the resuits that were presented here, there are some potential objections that need to be pointed out o 164 The next sub-sections examine four potential objections or qualifications. In particular, the research project’s retrospetive perspective; ambiguities with the central concept of “non-mainstream party;” the choice of attitude measures; and the consequences of non mainstream voting. Ail of these points require some further discussion. • Retrospectzue us. Prospectiue The project examined voting behaviour through retrospective lens, whereby voting behaviour is explained by using past economic data. On tlie surface, there may not need to be any justification, simply because one’s knowledge of his or lier long-term history is expected to be far better than forecasts of the long-term future. But on the other hand, voters do have expectations, and if long-term economic dechne leads voters to support a non-mainstream party, then certainly they have in mmd some idea of wliat such a party could do in the future. In a simiiar vein, support for any mainstream party may partly reftect that party’s stand on how to manage the economy over the Iong-term. Voters can, therefore, align their iong-term economic interests to an appropriate p arty. There is a possibihty for this type of voting. In the short-term, Nadeau and Lewis-Beck (2001) point out how in American presidential elections voters evaluate the potential future economic prospects of 165 candidates when neither is seeking re-election. There is nothing to suggest voters do not extend their evaluations into the long-term. After ail, it is rational to consider the long-term, sinc manv important decisions in life, such as assuming a mortgage to purchase a house, are not only economic, but require a long-term commitment and a sufficient level of economic health over that long period of time. The problem with a long-term prospective voter is that it is difficuit to measure. One’s support for a non-mainstream party can reflect future expectations of what that party can do. If a partv seems favourable to a particular sector, then voters who identify with that sector can decide whether to support that party. But there are at two problems. First, ail this assumes that a party delivers what it promises. Also, even if a party does deliver on its promises, there remains the distinct possibiiity that the policy ideas would nonetheless fail. Just because a party promises to enact certain measures aimed to help a certain sector does not automaticaliy imply that those measures will work, or continue to work, in the long run. Ail this makes it liard to build a dataset that matches a voter’s choice in one election witli the expected long-term economic performance that a particular party. Second, the only parties that can be judged on their ability to enact appropriate economic policy are those that form governments. Non-mainstream parties generally sit in opposition, and generally expect to remain in opposition. Therefore, they cari make ail kinds of promises 166 and daims without fear that crie day they will actually have to lie accountable to them. Voters mav lie cognizant of sucli a constraint. Instead of seeing sucli a vote as a waste, support for nonmainstream parties might be used te send a signal to the wining mainstream partv that something is amiss, and that future election gains depend on that partv’s appropriate reaction to redress long-term grievances. In sum, a vote for a non-mainstream party may reflect future expectations of what that non-mainstream party promises te deliver, or used te pressure the mainstream winning party te produce the same desired resuits. in anv case, a prospective approacli is difficuit te apply. That is why the retrospective approach applied here is seen as more feasible. It aise more closely tics into an analysis of how voters’ attitudes have evolved over time te reflect their dwindhng economic conditions, since a main point of this research is that support for non-mainstream parties reflects a general decline in attitudes towards the politicai system, and this in turn was caused by disappointing conditions over flic long-term. Ambiguity of the ‘Won-Mainstrenm” Concept The concept of a non-mainstream party is clear te the Canadian case, at least up until the 2000 generaÏ election. No other partv other than the Liberals and Progressive Conservatives lias ever governed 167 Canada. Ail other parties have done n better than to sit on the opposition benches. Even in minority governments, none of the other parties ever formed part of a coalitïon government. As mentioned previously, Canada produced oniy one coaiition government (1917- 1920), and even that was formed between Conservative and LiberaÏ MPs. Certainlv, the NDP has functioned as a major plaver, boistering Liberai minority governments in exchange for certain policv provisions. But even here, the NDP remained an opposition party. Aside for some bragging rights of having forced a Liberai government to pass this or that budget provision or legislative bili, and a moderatelv successful history of winning some eiections at the provinciai level, the NDP cannot ciaim to have a hïstory as a governing party in Canadian federai politics. They remain a non-mainstream party that cannot even to daim to have ever formed the officiai opposition. But since the 2004 election, the unes between non-mainstream and mainstream have blurred. The PC-Alliance merger mav be regarded as a blend of both mainstream and non-mainstream parties. So at best, the new Conservative party is a meso-mainstream party. However, an argument can be made that the new Conservative party is comprised mostly of the old Reform/Ailiance party, with far less of it comprised of whatever was left of the old PC party. So it is flot obvious how to interpret a vote for the Conservative party. Is it a vote for a non mainstream party, and therefore, attractive to those who feei that the political svstem needs fundamental change? Or is the new Conservative 168 partv more mainstream, attracting votes for those who wish to see more stabilitv? Is it as mainstream as the Liberals? Also, and this was pointed out in the conclusion of Article 1, the concept of non-mainstream employed here may no easily appiy elsewhere, particularly in countries whose electoral systems are far more proportional, and consequently produce multi-party systems. Often, governments are led by coalitions of two or more parties. Aside from the more extremist fringe parties that are almost aiways in opposition, some coalition members are parties that can be considered “radical.” For example, the ultra-orthodox party Shas was a member of Yitzhak Rabin’s coalition government in Israel in 1992; the anti-immigrant populist rightwing Austrian Freedom Party was a junior member of a coalition government led by the conservative Austrian People’s Party in 2000. In Canada’s plurality system, sucli parties would not stand a chance of becoming part of any coalition. Perhaps the concept should not be construed simply as a dichotomy, but more in terms of degree. A party then can be placed along an ordinal continuum that includes intermediate levels of mainstreamness. At the more “mainstream” end of the continuum would include such parties as the Liberal Party of Canada, the Republican and Democratic Parties of the United States and the Social Democrats in Germany. The opposite end of the continuum would include clearly non mainstream parties such as the Bloc Québécois, Ralph Nader’s Green Party, the British National Party, and the Japanese Communist Party. 169 The more grev middle area would include those parties that have formed part of a ruling coalition, aithougli as a junior member. Perhaps one level below that would include those parties that have not officially become part of a coalition, but have nonetheless provided support for a government, such as Canada’s NDP and Denmark’s People’s party.3 A revised conception cf “non-mainstream” should also take into account a party’s historv. There was a time in Britain when the Liberal party was the mainstream competitor to the Conservatives. Now, the Liberal party is a third party at best, and Labour has, since 1945, won many general elections, and now ranks with the Conservatives as Britain’s two main parties. How should the Liberals be classified? Does a long period of absence from power disqualify a party as mainstream? How long should a party have to sit in opposition hefore its “mainstream” status expires? Clearly, the concept “non-mainstream” cculd be elaborated further. Perhaps a mathematical formula could be designed to quantify a party’s level of mainstreamness by taking into account certain key factors. such as the party’s proportion of cabinet seats, number of coalition partners, tenure in office, and duration in opposition. In sum, the concept of non mainstream might need some more careful revision. Castles and Mair (1984) cornes close. They score parties using a left-right scale that ranges from O (extreme left) to 10 (extreme right). This may seem appropriate for most cases, but it can lead to an interpretation cf parties which is too focused on ideological distance as opposed to an insider-outsider status. for example, their results show that a score of 2. 1 separates the NDP (3.2) and the Liberals (5.3), while the distance between what is clearlv a more non mainstream party, the Social Credit (7.8), and the Progressive Conservatives (6.5) party is only 1.3. Under this scheme, Social Credit is doser to the mainstream than the NDP, a conclusion that is debatable. 170 • Attitude Measures Political efficacy fur:ctioned as a key factor in explaining why long-term economic decline raises the propensitv to vote non mainstream. As the summary resuits show, the inverse relationship between long-term economic decline and political efficacy is very strong. The longer one suffers economic decline, the more the political system will be regarded as unresponsive, and consequently, the more likely one begins to manifest lower levels of external political efficacy. But is political efficacy the only attitude that mediates the relationship between eroding conditions and voter support for non-mainstream parties? Another potentially relevant attitude is “trust.” Definitions of “trust” vary, but in general, trust is an evaluative attitude about individuals or administrations. Political trust in particular is a belief that political leaders can lie expected to render good governance. If is an index of survey items that refer to incumbents, leaders (i.e., actual people), or a particular administration (e.g., party in power). for example, one item is as follows: “Do you think the people in government waste a lot of the money we pay in taxes, waste some of it, or don’t waste very much of it?” The referent “people in government” conjures mental pictures of the head of government, such as the prime 4 Both the American National Election Study and the Canadian Election Study use similarly worded items. 171 minister, key cabinet members, or more generic images of politicians or civil servants. In any case, the item points to people. The concept has recently drawn a considerable amount of attention in comparative political analysis, particularly in trying to explain why citizens vote less and appear to be more cynical towards the political establishment, especially since the 1970s (see, for example, Hardin, 2000; Newton, 1999; Newton and Norris, 2000). One recent Canadian study (Bélanger and Nadeau, 2005) points to an inverse and consistent relationship between trust and support for third parties (NDP, Reform and Bloc Québécois), while political trust appears to yield almost no impact on voter support for the mainstream opposition party. On the surface, one could draw parallels between declines in political trust and declines in political efficacy, both in Canada and abroad. Both trends seem to have taken place over the same economically turbulent period, and both have led to similar results: the rise of non-mainstream parties, voter cynicism and lower turnout. But political trust and political efficacy are different concepts. They mav correlate very well, but they measure different phenomenon, and yield different implications. Trust is a central idea to the more broad study of social capital, which has also been on the decline, but not necessarily because of economic decline. According to Putnam (2000), the decline of social capital is a result of older cohorts, who were raised during a period wlien social engagement (especially through church groups) was more 172 commonplace (or expected), being replaced b1,’ younger cohorts, who lived under a differer social context, such as suburbanization, the aUvent of television and secularization. The decline of social engagement lias therefore leU to a decline in overali levels of trust, since, as Newton (1999) points out, trust is something that strengthens with interpersonal contact (i.e., engagement). Therefore, the decline in social capital lias constrained the development of any form of public trust. In sum, this phenomenon’s relationship with economic decline may only be coincidental, and seems more tied to sociological, not structural, changes. Another problem with political trust is that it may not necessarily function as a valid measure of system-wide disaffection. As mentioned before, indicators of trust measure evaluations of particular leaders, incumbents or administrations. They do not primarily focus on the extent to which the political system as a whole is responsive to its citizens. Trust items may not strictly measure “specific” political support, but they tap info evaluations of political actors rather than the overail regime. Also, one reason why measures of trust capture some system-wide sentiments is because citizens may naturally progress from having a sense of mistrust for particular leaders into a more generalized rejection of the political system. As Hetherington points out: “As problems go unsolved over a series of administrations, citizens may begin to question the regime,” (1998: 792). In other words, as trust for political actors continues to dwindle, negative views of the political 173 svstem will eventually follow. Put more simply: specific measures of trust and the more system-wide evaluations correlate over time. Aiso, since trust items refer to more specific political actors, negative trust cari be addressed througli regular elections. As Abramson and Finifter (1981) point out: “Disaffection with incumbents can, in principie at least, be remedied through the electoral process,” (298). Simply eÏect new leaders, or vote for the mainstream opposition. But the propensity to support non-mainstream parties benefits from negative evaluations of the entire political svstem, not just particular leaders. Even the mainstream officiai opposition is seen as dysfunctional or unable to exact adequate change. However, it must be acknowledged that non-mainstream parties do not only focus their criticism against fric system. They do name names and do point out how this or that leader lias failed the public. Such an approacli renders relevant the concept of political trust. However, a large part of a non-mainstream party’s raison d’étre is to challenge poÏitics as usuai, in its entirety. Tliey criticize the overail political system’s inability to respond effectively to citizen demands. In this light, external political efficacy seems more appropriate, since it is a more specffic measured of support for the entire political svstem, and not any particular administration or particular incumbents. o 174 • Why vote non-mwnstrearn? Another potential objection concerns an omission. Although the link between long-term economic decline and non-mainstream voting has been establislied, and altliough this is tied to political efficacv, one central question remains unanswered: What benefits might a voter derive from supporting a party that clearly could do no better than win enough seats to form a viable opposition party, and at worse, shift votes away from a second-choice contender thereby allowing a lesser preferred third-choice candidate to win. This question is centered flot su much on wliat motivates voters to choose non-mainstream parties, but what satisfaction do they get from having done so? As Anderson et al. (2005) point out, voters who supported parties that did not win an election tend to have less favourable evaluations of the electoral process. This is truc in general, and includes parties that might otherwise have won an election (i.e., the mainstream opposition). But liow do voters evaluate the democratic process after knowingly voting for a losing non mainstream party? Worse stili (and this applies particularly to Canada’s plurality system), lio does a voter respond after having cast a vote for an un-winnable non-mainstream party, and then seeing bis or lier Ïeast preferred candidate xvin the constituency? These questions are not just academic. As sliown in Figure 1. 1 in the flrst article, voting in Canada lias shown remarkable trends in favour of non-mainstream parties that have no realistic chance of 175 forming a government. If indeed the survival of a democracy depends on the consent of the 1oser then the survival of Canada’s democracv depends not only on that almost 60 per cent of the electorate that does not cast a vote for winner party (since a party tvpically needs littie more than 40 per cent of the vote in order to form a majority government), but also on that 25 to 45 per cent of voters who over the course of the last few decades have supported parties that are consistent losers. If there is any legitimacv crisis that will erupt in Canadian politics, it may emerge from that portion of the Canadian electorate that considers itself a consistent loser both at the ballot box and in the economic agenda. 4- Final Remarks Some general concluding remarks can be made bv drawing an anaïogy with a completely different field. In the world of consumer behaviour, a change in consumer tastes in the market often leads to a demand shock. This is natural, and expected, although the shock can be costly to the provider of a good or service that lias gone out of style. But in the marketplace of political ideas, the consequences of demand sliocks are not aiways benign, and almost neyer just local. When voters change their “tastes,” it may lead the emergence of new political parties. But when that change stems from discontent, then the context becomes less stable and potentially more dangerous. Voter discontent suggests a 176 failure of the regime, whicli opens the door to both potendallv progressive ideas (democratization, reform, policy innovation, etc.), and also reactionary and potentially authoritarian movements. It is not necessarily predictable which way society xviii go. It is also n)t easy to identify the beginnings of such trends. Typically, ail the attention in an election is given to the largest parties. Smailer parties liardly get any media coverage. It is oniy when non mainstream parties begin to win seats or begin to measure high in public opinion polis that the presence and growth of such movemcnts get noticed. And if it takes voters a long period of time to develop a taste for non-mainstream parties, then the reversai may require at ieast as long a period of time. If a growing non-mainstream party threatens politicallv stabuiity or even liberai democratic principies, then surely one should not sit idly by and wait for this movement to simply “fizzle out.” furthermore, the fact that some of their supporters are driven by sentiments that suggest disapprovai of tlie entire politicai system imposes a certain politicai agenda. Voters cleariy expect some things to get done, and untii they see a long-term record of positive results, thev are not likeiy to return to mainstream parties. So what’s a government to do? First, one should not lie led to beiieve that the growth of non-mai’stream parties necessarily impiies a mardi down towards a revoiution. That is definiteiy a possibulity, and history shows that societies can turn against regimes that fail to 177 respond appropriately to pressing problems. But Canada may not be a good case for regime failure. Non-mainstream parties are not necessarily destabilizing or violent. Some aspects of the policy platforrns of non mainstream parties may be innovative, some even radical, and followers (and sornetimes even leaders) of non-mainstream parties spew out the occasional intolerant rernark. But in general, the Canadian political system appears sufficiently responsive. The centrist Liberal party has a historv of leaning a littie to the left and a littie to right, depending on the political winds. Also, Reform/Alliance recognized that greater electoral success can only corne about by shifting doser towards the mainstream, which included a merger with whatever remained of the Progressive Conservative party. InstitutionaÏly, then, Canada appears to have shock absorbers in place, and the Canadian political culture appears to dernand centrist parties that make every effort possible to behave “mainstream.” But if the rise of electoral support for non-mainstream parties is indeed tied to voter discontent with the political systern as a whole, then one obvious solution is to reforrn the political system. But this solution is not aiways feasible. Any political reform that requires constitutional arnendments is off the table. Even non-constitutional reform, such as reforming the electoral system, is proving difficuit to accomplish. So, for the foreseeable future, don’t bet on institutional reform to cure political ilis. o 178 But one thing that governments can do, and sometimes do well, is put in place policies that encourage more equitable economic conditions. It is not to say that the Canadian government should embrace socialism, but if there is a group that is suffering long-term economic decline, then the state can step in. Either the state can draft policies to reverse a decline, or, if the decline is seen as necessary as part of an industrial evolutionary process (e.g., the rise of the automotive industrv implied a decline in occupations such as ferriers, whip makers, stagecoacli builders, etc.), then programs can be put in place to help retrain and reintegrate those being left out by economic change. Healtliy macroeconomic management is the least that a government can pursue, and if it does that well, it may minimize the strength of potentiallv destabilizing political movements. Furthermore, institutional reform may be facilitated by positive economic conditions. In other words, future political and institutional reform to improve Canada’s democratic and federal system may require first that the state ensure that Canadians generally feel that their economic needs are satisfied. Or, minimally, citizens who struggie economically must not be given any reason to feel that the state lias abandoned tlieir pliglit. o 17Ç) 5IBLIoGIAPHY Abramson, Paul R., and John H. Aldrich. (1982). “The Decline of Electoral Participation in America.” American Politic& Science Review76: 502-21. Abramson, Paul. R., and Ada W. Finifter. (1981). “On the Meaning of Political Trust: New Evidence from Items Introduced in 197$.” American Journal ofPolitical Science 25: 297-307. Aish, Anne-Marie, and Karl G. Jôreskog. (1990). “A Panel Model for Political Efficacy and Responsiveness: An Application of LISREL 7 with Weighted Least Squares.” Quatity & Quantity 24: 405-26. Aimond, Gabriel, and Sidney Verba. (1963). Civic Culture. Boston, MA: Littie, Brown & Company. Anderson, Christopher J. (2000). “Economic Voting and Politicai Context: A Comparative Perspective.” Electorat Studies 19: 151- 170. Anderson, Christopher J., André Biais, Shaun Bowler, Todd Donovan and Ola Listhaug. (2005). Losers’ Consent: Etections and Democratic Legitimacy. Oxford: Oxford University Press. Balch, George. (1974). “Multiple Indicators in Survey Research: The Concept ‘Sense of Poiitical Efficacy.” Political MethodoÏogy 1: 1- 43. Bale, Tim. (2003). “Cinderella and Her Ugly Sisters: The Mainstream and Extreme Riglit in Europe’s Bipolarising Party Systems.” West European Politics 26: 67-90. Bélanger, iric. (2004). “Anti-Partyism and Third-Party Vote Choice: A Comparison of Canada, Britain, and Australia.” Comparative Political Studies 37: 1054-1078. Bélanger, ric, and Richard Nadeau. (2005). “Political Trust and the Vote in Multiparty Elections: The Canadian Case.” European Journal of Potiticat Research. 44: 121-146. Betz, Hans-Georg, and Carol Jolinson. (2004). “Against the Current Stemming the Tide: The Nostalgic Ideology of the Contemporary Ideologies 9: 311-327. o Radical Populist Right. Journal of Potitical 180 Biais, André. (2005). “Accounting for the Electoral Success of the Liberal Partv in Canada.” Presidential Address to the Canadian PoÏitical Science Association, 3 June 2005, London ON. Biais, André, Elisabeth Gidengii, Richard Nadeau and Neii Nevitte. (2002). Anatomy of n Liberal Victory. Peterborough, ON: Broadview Press. Blishen, Bernard R., William Carroli and Catherine Moore. (1987). “The 1981 Socioeconomic Index for Occupations in Canada.” Canadian Review of SocioÏoyy and Anthropologtj 24: 465-88. Brady, David. (2003). “Rethinking the Sociological Measurement of Poverty.” Soctat forces 81: 715-52. Brady, David W., Halirie C. Han, and Jeremy C. Pope. (2005). “Primary Elections and Candidate Ideology: Out-of-Step with the Primary Electorate?” Unpublished working paper. Brooks, Ciem, and David Brady. (1999). “Income, Economic Voting, and Long-Term Political Change in the U.S., 1952-1996.” Sociat forces 77: 1339-1375. Burnside, Ross, Stephen Herbert, and Stephen Curtis. (2003). Election 2003. Edinburgh: Scottisli Parliament Information Centre. Campbell, Angus, Gerald Gurin and Warren E. Miller. (1954). The Voter Decides. Evanston, IL: Row Peterson. Carisen, Fredrick. (2000). “Unemployment, Inflation, and Government Popularity — Are There Partisan Effects?” EÏectoral Studies 19: 141-150. Casties, Francis G., and Peter Mair. (1984). “Left-Right Pohtical Scales: Some ‘Expert’ Judgments.” European Journal of Politicat Research 12: 73-88. Chappeli, Henrv W., and Linda Gonçalves Viega. (2000). “Economics and Eiections in Western Europe: 1960-1997.” Electoral Studies 19: 183-197. Clarke, Harold D., and Alian Kornberg. (1993). “Evaluations and Evolution: Public Attitudes Toward Canada’s Federai Pohtical Parties, 1965-1991.” Canadian Journal ofPotiticaÏ Science 26: 287-311. o 1.9,1 Clarke, HarolU D., and Allan Xornherg. (1996). “Partisan Dealignment, Electoral Choice and Partv-Svstem Change in Canada.” Patty PoÏztics 2: 455-78. Clarke, Harold D., Allan Kornberg, Faron Ellis and Jon Rapkin. (2000). “Not for fame or Fortune.” Party PoÏitics 6: 75-93. Craig, Stephen C., and Michael A. Maggiotto. (1982). “Measuring Political Efficacy.” Politicat Methodology 8: 85-109. Craig, Stephen C., Richard G. Niemi and Glenn E. Silver. (1990). “Political Efficacy and Trust: A Report on the NES Pilot Study Items.” PoliticaÏ Behauiour 12: 289-3 14. Crompton, Susan, and Michael Vickers. (2000). “One Hundred Years of Labour Force.” Canadian Social Trends 57: 2-13. Cutier, Fred. (2002). “Local Economies, Local Policy Impacts and Federal Electoral Behaviour in Canada.” Canadian Journal of Political Science 35: 347-82. Downs, Anthony. (1957). An Economic Theory ofDemocraccj. New York: Harper and Row. Duverger, Maurice. (1954). Political Parties. London: Methuen. Easton, David. (1975). “A Re-Assessment of the Concept ofPolitical Support.” British Journal of Political Science 5: 435-57. Easton, David, and Jack Dennis. (1967). “The Chuld’s Acquisition of Regime Norms: Political Efficacy.” American Politicat Science Reuiew6l: 25-38. Erikson, Robert S., Michael B. MacKuen and James A. Stimson. (2000). “Bankers or Peasants Revisited: Economic Expectations and Presidential Approval.” ElectoraÏ Studies 19: 295-3 12. Feidman, Stanley. (1982). “Economic Self-Interest and Political Behaviour.” Arnerican Journal of PoÏiticaÏ Science 26: 446-466. Feidman, Stanley. (1985). “Economic Self-Interest and the Vote: Evidence and Meaning.” In Heinz Eulau and Michael S. Lewis Beck (Eds). Economic Conditions and Electoral Outcomes: The United States and Western Europe. New York: Agathon Press. Fieschi, Catherine, and Paul Heywood. (2004). “Trust, Cynicism and Populist Anti-Politics. Journal of Political Ideologies 9: 289-309. 182 Fiorina, Morris. (1978). “Economic Retrospective Voting in American National Elections: A Micro-Analvsis.” American JournaÏ of PoÏiticat Science 22: 426-443. Gidengil, Elisabeth. (1990). “Centres and Peripheries: The Political Culture of Dependenc, .“ Canadian Review ofSocioÏogy and AnthropoÏogy 27: 23-48. Gidengil, Elisabeth André Biais, Richard Nadeau and Neil Nevitte. (1999). “Making Sense of Regional Voting in the 1997 Canadian Federal Election: Liberal and Reform Support Outside Quebec.” Canadian Journal of Political Science 32: 247-72 Gidengil, Elisabeth , André Blais, Neil Nevitte and Richard Nadeau. (2001). “The Correlates and Consequences of Anti-Partyism in the 1997 Canadian Election.” Party Politics 7: 491-513. Givens, Terri E. (2002). “The Role of Socioeconomic Variables in the Success of Radical-right parties.” In Shadows Over Europe: The Devetopment and Impact ofthe Extreme Right in Western Europe, eds. Marsin Schain, Aristide Zolbert and Patrick Hossay. Hampshire, England: Palgrave MacMillan. Godbout, Jean-François, and ric Bélanger. (2002). “La dimension régionale du vote économique canadien aux élections fédérales de 1988 à 2000.” Canudian Journal of PoÏitical Science 35: 567- 88. Golder, Matt. (2003). “Expiaining Variation in the Success of Extreme Right Parties in Western Europe.” Comparative Political Studies 36: 432-466. Goodhart, C.A.E., and R.J. Bhansali. (1970). “Politic’i Economy.” Political Studies 18: 43-106. Gunther, Richard, and Larry Diamond. (2003). “Species of Political Parties.” Party Politics 9: 167-99. Hardin, Russell. (2000). “The Public Trust.” In Disaffected Democracies: What’s Troubling the Trilateral Countries? eds., Susan J. Pharr and Robert D. Putnam. Princeton: Princeton University Press. Hearl, Derek J., Jan Budge, and Bernard Pearson. (1996). “Distinctiveness of Regional Voting: A Comparative Analysis Across the European Community (1979-1993).” Electoral Studies 15: 167-182. 183 Henderson, Ailsa. (2004). “RegionaÏ Political Cultures in Canada.” Canadian Journal of PoÏitical Science 37: 595-6 15. Hetherington, Marc J. (1998). “The Political Relevance of Political Trust.” American PoÏiticaÏ Science Rer’iew 92: 79 1-808. Hetliermgton, Marc J. (1999). “The Effect of Political Trust on the Presidential Vote, 1968-96.” American PoÏitical Science Reuiew 93: 311-326. Heisz, Andrew, Andrew Jackson and Garnett Picot. (2002). “Winner and Losers in the Labour Market ofthe 1990s.” Statistics Canada. Catalogue No. 11F0019M1E2002184. Hibbs, Douglas A. (1987). Hie American Political Economy. Cambridge, MA: Harvard University Press. Hoizer, Harry J. (1990). “Labour Force Participation and Employment Among Young Ment Trends, Causes, and Pohcy Implications.” Research in LabourEconomics. 11: 115-36. Ignazi, Piero. (2003). Extreme Right Parties in Western Europe. Oxford: Oxford University Press. Iyengar, Shanto. (1980). “Subjective Political Efficacy as a Measure of Diffuse Support.” Public Opinion Quarterly 44: 249-56. Jackman, Robert W., and Karin Voipert. (1996). “Conditions Favouring Parties of the Extreme Right in Western Europe.” Rriîish JournaÏ ofPoÏitical Science 26: 50 1-521. Jackson, Robert J., ami Doreen Jackson. (2006). Politics in Canada: Culture, Institutions, Behaviour and Public Poticy. Toronto: Pearson Education Canada. Johnston, Richard, André Biais, Henry E. Brady and Jean Crête. (1992). Letting the People Decide: Dynamics of n Canadian Election. Montreal, QC: McGill-Queen’s Press. JoIly, Seth Kincaid. (2004). “European Integration and the Rise of Regional Parties.” Prepared for the “Democracy in Europe: European Politics and Multi-Level Governance Summer School, 12-24 September 2004, Vrije Universiteit, Amsterdam. Kev, VO. (1966). The Responsible Electorate. Cambridge, MA: Harvard o University Press. 184 Kinder, Donald R., and D. Roderick Kiewiet. (1979). “Economic Discontent and Political Behaviour: The Role of Personal Grievances and Collective Economic Judgements in Congressional Voting.” American Journal ofPotitical Science 23: 495-527. Kinder, Donald R., and Roderick Kiewiet. (1981). “Sociotropic Politics: The American Case.” British Journal ofPolitical Science 11: 129 61. King, Gary, Michael Tomz and Jason Wittenberg. (2000). “Making the Most of Statistical Analyses: Improving Interpretation and Presentation.” American Journal 0f Potitical Science 44: 34 1-55. Kramer, Gerald H. (1971). “Short-Term Fluctuations in US. Voting Behaviour, 1896-1964.” American Political Science Review 65: 131-143. Kramer, Gerald H. (1983). “The Ecological fallacy Revisited: Aggregate Versus Individual-Level Findings on Economics and Elections, and Sociotropic Voting.” American Political Science Reuiew 77: 92-111. Krause, George A. (1997). “Voters, Information Heterogeneity, and the Dynamics of Aggregate Economic Expectations.” Arnerican Journal of Political Science 41: 1170-200. Landes, Ronald. (2002). The Canadian PoÏity. Toronto: Prentice Hall. Lane, Robert E. (1959). Political Life: Why and How People Get Inuolued in Politics. New York, NY: The Free Press. Lewis-Beck, Micliael S. (1986). “Comparative Economic Voting: Britain, France, Germany and Italy.” American Journal ofPotiticat Science 30: 315-346. Lewis-Beci Michael S. (1988). Economics and Elections: The Major Western Democracies. Ann Arbor: University of Michigan Press. Lewis-Beck, Michael S., and Martin Paldam. (2000). “Economic Voting: An Introduction.” Electoral Studies 19: 113-21. Lewis-Beck, Michael S., and Charles Tien. (2004). “Jobs and the Job of President: A Forecast for 2004.” PS: PoliticuÏ Science and Politics 37: 753-58. 185 Lipset, Seymour Martin. (1959/63). Politicat Man. Garden City: Anchor Books, Doubleday & Company. MacKuen. Michael B., Robert S. Erikson and James A. Stimson. (1992). “Peasants or Bankers? The American Electorate and the US. Economy.” American Potitical Science Review 86: 597-6 11. Madsen, Douglas. (1978). “A Structural Approach to the Explanation of Political Efficacy Levels Under Democratic Regimes.” American Journal ofPoÏiticaÏ Science 22: $67-83. Mair, Peter. (2001). “The Green Challenges and Political Competition: How Typical is the German Experience?” German PoÏitics 10: 99- 116. Meguid, Bonnie M. (2005). “Competition Between Unequals: The Role of Mainstream Party Strategy in Niche Party Success.” American PoliticaÏ Science Reuiew 99: 347-359. Miiler, Warren E., and J. Merriil Shanks. (1996). The NewAmerican Voter. Cambridge, MA: Harvard University Press. Mondak, Jeffery J., Diana C. Mutz and Robert Huckfeldt. (1996). “Persuasion in Context: The Muitilevel Structure of Economic Evaluations.” In Potiticat Persuasion and Attitude Change, cd. Diana C. Mutz, Paul M. Sniderman, and Richard Brody. Ann Arbor: University of Michigan Press. Morton, Wiiliam L. 1950. The Progressive Party in Canada. Toronto: University of Toronto Press. Mughan, Anthony, and Dean Lacy. 2002. “Economic Performance, Job Insecurity and Electoral Choice.” British Journal of PotiticaÏ Science 32 (Juiy): 5 13-533. Nadeau, Richard, and André Biais. (1993). “Expiaining Election Outcomes in Canada: Economy and Politics.” Canadian Journal ofPotiticat Science 26: 775-790. Nadeau, Richard, and André Biais. (1995). “Economic Conditions, Leader Evaluations and Eiection Outcomes in Canada.” Canadian Public Policy 21: 2 12-218. Nadeau, Richard, André Biais, Neil Nevitte and Elizabeth Gidengil. (2000). “Its Unemployment, Stupid! Why Perceptions About the Job Situation Hurt the Liberais in the 1997 Eiection.” Canadian O Public Poticy 26: 77-97. 186 Nadeau, Richard, and Michael S. Lewis-Beck. (2001). “National Economic Voting in US. Presidential Elections.” Journal of PoÏitics 63: 159-8 1. Nagler, Jonathan, and Suzanne De Boef. (2000). “The Varying Self Interest of Voters and Economic Reference Groups- Two Analyses to Explain Al Gore’s 2000 Campaign Strategy.” Paper presented to the American Political Science annual meeting, Washington, DC. Nagler, Jonathan, and Jennifer R. Niemann. (1997). “Economic Conditions and Presidential Elections.” Paper presented to the Annual Meeting of the American Political Science Association, Washington, DC. Nagler, Jonathan, and Jennifer R. Willette. (1999). “Economic Conditions and Presidential Elections.” Barcelona: Institut de Ciències Politiques i Socials, Barcelona. Working Paper No. 171. Nannestad, Peter, and Martin PaÏdam. (1994). “The VP-Eunction: A Survey of the Literature on Vote and Popularitv Functions After 25 Years.” Public Choice 79: 213-245. Nannestad, Peter, and Martin Paldam. (1995). “It’s the Government’s Fault! A Cross-Section Study of Economic Voting in Denmark, 1990/93.” European Journal ofPotitical Research 28: 33-62. Nannestad, Peter, and Martin Paldam. (1997). “From the Pocketbook of the Welfare Man: A Pooled Cross-Section Study of Economic Voting in Denmark, 1986-92.” British Journal ofPoÏiticaÏ Science 27: 119-136. Norris, Pippa. (2005). Radical Right: Voters and Parties in the Regulated Market. New York: Cambridge University Press. Newton, Kenneth. (1999). “Social and Political Trust in Established Democracies.” In Critical Citizens: Global Support for Democratic Governance, ed. Pippa Norris. Oxford: Oxford University Press. Newton, Kenneth, and Pippa Norris. (2000). “Confidence in Public Institutions: Faith, Culture, or Performance?” In Disaffected Democracies: What’s Troubhng the Trilateral Countries? eds., Susan J. Pharr and Robert D. Putnam. Princeton: Princeton University Press. Ostry, Sylvia, and Mahmood A. Zaidi. (1979). Labour Economics in Canada. Toronto: Macmillan of Canada. 187 Parsons, Donald 0. (1980). “T1e Decline in Male Labour Force Participation.” MonthÏy Labour Review. february, 117-34. Pharr, Susan J., and Robert D. Putnam. (2000). Disaffected Democracies: What’s TroubÏing the Trilateral Countries? Princeton: Princeton University Press. Picot, Garnett, and Andrew Heisz. (2000). “The Performance ofthe 1990s Canadian Labour Market.” Canadian Pubtic Policy 26: S7- S2 5. Pinard, Maurice. (1971). The Rise ofa Third Parfy. Englewood Cliffs, NJ: Prentice-Hall. Pinard, Maurice. (1973). “Third Parties in Canada Revisited: A Rejoinder and Elaboration to the Theory of One-Party Dominance.” Canadian Journal ofPolitical Science 6: 439-60. Powell, G. Bingham, and Guy D. Whitten. (1993). “A Cross-National Analysis of Economic Voting: Taking Account of the Political Context.” American Journal of Politicat Science 37: 39 1-414. Poguntke, Thomas. (1996). “Anti-Party Sentiment — Conceptual Thoughts and Empirical Evidence: Explorations into a Minefield.” European Journal ofPolitical Research 29: 319-44. Putnam, Robert. (2005). Bowling Atone: The Collapse and Reuivat of American Community. New York: Simon and Schuster. Rattinger, Hans. (1991). “Unemployment and Elections in West Germany.” In: Helmut Norpotli, et al., (Eds.) Economics and PoÏitics: The Calculus of Support. Ann Arbor: University of Michigan Press, pp. 49-62. Romero, David W., and Stephen J. Stambough. (1996). “Personal Economic Well-Being and the Individual Vote for Congress: A Pooled Analysis, 1980-1990.” Political Researrh Quarterly 49: 607-616. Sapsford, David. (198 1). Labour Market Economics. London: George Allen & Unwin Publishers Ltd. Sartori, Giovanni. (1976). Parties and Party Systems. New York: Cambridge University Press. 188 Sirneon, Richard, and David J. Elkins. (1974). “Regional Political Cultures in Canada.” Canadian Journal of Potitical Science 7: 397- 437. Smelser, Neil J. (1963). Theory of Collective Behaviour. New York: free Press of Glencoe. Smiley, Donald V. (1971). “The Structural Problem ofCanadian Federalism.” Canadian Public Administration 14: 326-43. Sniderman, Paul M., and Richard A. Brody. (1977). “Coping: The Ethic of Seif-Reliance.” American Journal ofPolitical Science 21: 501- 522. Southwell, Priscilla L. 1996. “Economic Salience and Differential Abstention in Presidential Elections.” American Politics Quarterly 24 (April): 22 1-236. Southwell, Priscilla L., and Marcy Jean Everest. 1998. “The Electoral Consequences ofAlienation: Nonvoting and Protest Voting in the 1992 Presidential Race.” Social Science Journal 35 (1): 43-8 1. Statistics Canada. (Various years). “Labour Force Survey.” CANSIM Tables No. 282-0002, 379-0017. Ottawa: Statistics Canada. Statistics Canada (1981). “Census of Canada. Labour force — Occupations Trends.” Cat. No. 92-920. Ottawa: Statistics Canada. Statistics Canada. (1993). “Occupation.” Cat. No. 93-327. Ottawa: Statistics Canada. Swank, Otto. (1993). “Popularity Functions Based on the Partisan Theory.” Public Choice 75: 339-356. Taggart, Paul. (1998). “A touchstone of dissent: Euroscepticism in Contemporary Western European Party Systems.” European Journal ofPolitical Research 33: 363-388. Taggart, Paul. (2004). “Populism and Representative Politics in Contemporary Europe.” Journal of Political Ideologies 9: 269-288. Teixeira, Ruy. (1998). “The New Economics of Voting.” Challenge 41: 19-37. Teixeira, Ruy A., and Joel Rogers. (2000). America’s forgotten Majority: Why the Vvhite Working Class StitÏ Matters. New York: Basic O Books. 189 Tomz, Michael, Jason Wittenbe:-g and Gary Ring. (2003). “CLARIFY: Software for Interpreting and Presenting Statistical ResuÏts. Version 2.1.” Stanford University, University ofWisconsin, and Harvard University. January 5. Available at http: / / gking.harvard.edu/ Van Der Brug, Wouter, Cees Van Der Eijk and Mark Franklin. (2000). “The Economy and the Vote: Electoral Responses to Economic Conditions in 15 Countries.” Paper presented to the American Political Science annual meeting, Washington, DC. Weatherford, M. Stephen. (1978). “Economic Conditions and Electoral Outcomes: Class Differences in the Political Response to Recession.” Amercan Journal of Politicat Scieace 22: 9 17-38. Wu, Chung-Li. (2003). “Psycho-Political Correlates of Political Efficacy: The Case of the 1994 New Orleans Mayoral Election.” Journal of Black Studies 33: 729-760. o 190 APPENifix 1: Outiine of Variables for Article 1 The units in the dataset used in Article 1 are provincial-level voting and economic data for seven elections years (1979, 1980, 1984, 1988, 1993, 1997 and 2000). for each eection, and for each province, the following was determined: Vote share for incumbent party: Provincia-1eve1 percentage of votes won by the ruling incumbent party in a general federal election. The incumbent was the Liberal Party in 1979, 1984, 1997 and 2000, and the Progressive Conservative Party in 1980, 1988 and 1993. Vote share for non-mainstream parties: Provincial-level percentage of votes won by parties other than the Liberal or Progressive Conservative Party, including independents. Long-term participation rate: The 10-year percentage change in provincial-level labour-force participation rate. The figure is calculated for each election year, as follows: P — P pH Ii) APL = 191 wher APL is the lO-vear percentage change in labour-force participation rate for a province (sub-script p) at election vear (sub script y); Pis the labour-force participation rate for a province p at year y; 10) is the labour-force participation rate for a province p at 10 vears before the general federal election held at year y. Long-term unemployment rate: The 10-year percentage change in provincial-level unemployment rate. The figure is calculated for each election year, as follows: — = Ufl AUL,, —I O) where UL1,0 is the percentage change in unemployment rate for a province (sub-script p) at election vear (sub-script y); is the unemployment rate for province p at year y; Up(y1o) is the unemployment rate rate for a province p at 10 years before the general federal election held at year y. Short-term participation rate: The one-year percentage change in provincial-level labour-force participation rate. The figure is calculated for each election year, as follows: 9 192 P -P APS — “ p ‘ptIt where APS is the percentage change in labour-force participation rate for a province (sub-scriptp) at election year (sub-script y); P, is the labour-force participation rate for province p at year y; P (l) is the labour-force participation rate for a province p one year years before the general ferlerai election held at vear y. Short-term unemployment rate: The 1O-year percentage change in provincial-level unemployment rate. The figure is calculated for cadi election year, as follows: t) A /‘ p) where AUSis the one-year percentage change in unemployment rate for a province (sub-script p) at election year (sub-script y); Upy is the unemployment rate for province p at year y; Up(y1) is the unemployment rate for a province p one year before the general ferlerai election helU at year y. Atlantfc, Quebec, West: Regional dummy variables. A score of 1 is assigned to a unit if the province beloncs to one of these regions. 193 Ontario lias been selected as the reference group. The northern territories have been excluded. o 194 N APPENDIx 2: Outlïne of Variables for Article 2 The units in the dataset used in Article 2 pooled together the 1979 and the 1993 Canadian Election Study. Vote choice: A categorical variable based on reported voting behaviour of respondents from the 1979 and the 1993 Canadian Election Studv (post-election wave). A score of 1 is assigned to respondents who reported voting for the incumbent (i.e., the Liberals in 1979; the Progressive Conservatives in 1993), a 2 if they reported a vote for a non- main stream party, and a 3 if they reported to have abstained from voting. 1O-year change in occupational growth rate: A respondent of a particular election year is assigned a value pertaining to the long-term economic change of lis or ber occupational group. The long-term change is based on the labour market data drawn from the Canada Census nearest to the election year and 10 years prior (i.e., respondents from the 1979 and 1993 CES are assigned values drawn from, respectively, the 1971 and 1981 Census, and the 1981 and 1991 Census). Negative values reflect shrinkage; positive values reflect o 195 gro\h. For purposes of analysis, positive values are set to zero, whule negative values are converted to positive scores. External political efficacy: An additive composite measure based on two CES questions: 1) “I don’t think that the government cares mucli what people like me think,” and 2) “People like me don’t have any say about what the government does.” For each questions, a score of O is assigned to those who answered “Strongly agree,” .33 to “Agree,” .66 to “Disagree,” and 1 to “Strongly disagree.” Non-response (e.g., don’t know, refused to answer, etc.) are factored out. The scores on each of these questions are added. The sum is then divided by 2 ïn order to bring the index to a O to 1 scaÏe. Atiantîc, Quebec, West: Regional dummy variables. A score of 1 is assigned to a respondent if he or she resides in one of these regions. Ontario has been selected as the reference group. The northern territories have been excluded. Religion: A dummy y iriable whereby respondents who are Catholic are assigned a score of 1, ail others a O. Language: A dummy variable whereby respondents who are francophone are assigned a score of 1, ail others a O. 196 O Age: A numeric variable that represents the age of each respondent. Gender: A dummv variable whereby a score of 1 is assigned to males, O to females. Union: A dummy variable whereby a score of 1 is assigned to respondents who have indicated membership to a labour union. 1993: A dummy variable that assigned as core of Ï to respondents from the 1993 Canadian Election Study. o 197 Q APPENDIx 3: Outline of Variables for Article 3 Datasets were compiled by pooling together Canadian Election Studv survevs from 1974 t 1997. Vote choice: A categorical variable based on reported voting behaviour of respondents from the 1974 to the 1997 Canadian EÏection Study (post-election wave). A score of 1 is assigned to respondents who reported voting for the incumbent (i.e., the Liberals in 1974, 1979, 1984 and 1997; the Progressive Conservatives in 1988 and 1993), a 2 if they reported a vote for a mainstream opposition partv, and a 3 if thev reported to have voted for a non-mainstream party. External political inefficacy: An additive composite measure based on two CES items: 1) “I don’t think that the government cares much what people like me think,” and 2) “People like me dont have any say about what the government does.” for each questions, a score of 1 is assigned to those wlio answered “Strongly agree,” .66 to “Agree,” .33 to “Disagree,” and O to “Strongly disagree.” Non-response (e.g., don’t know, refused to answer, etc.) are factored out. The scores on each of these questions are added. The sum is then divided by 2 in order to bring the index to a O to 1 scale. o 198 Economically declining demographic group: A dummv vahable that assigns a score of one to respondents who fuffihi the following three criteria: employed in a working-class occupation (forestrv, mining, processing, machining, fabricatio:i, construction, clerical, service, transportation), and have indicated an educational attainment level of below the post-secondary Ïevel (high-school diploma and less). Atiantic, Quebec, West: Regional dummy variables. A score of 1 is assigned to a respondent if he or she resides in one of these regions. Ontario bas been selected as the reference group. The northern territories have been excluded. Religion: A dummy variable whereby respondents wlio are Catholic are assigned a score of 1, ail others a O. Language: A dummv variable whereby respondents who are francophone are assigned a score of 1, ah others a O. Age: A numeric variable that represents the age of each respondent. Gender: A dummy variable whereby a score of 1 is assigned to males, O to females. o Union: A dummy variable whereby a score of 1 is assigned to respondents who have indicate1 membership to a labour union. 1979, 1984, 1988, 1993: Election-year dummy variables. A score of 1 is assigned to a respondent if lie or she lias participated in tlie election study of one of these years. C 200